Everything routerflow collects and measures, as tidy tables you can query: one Parquet file per table, and one DuckDB file holding them all. Rebuilt every hour on the collection server from the daily snapshots, the raw files and the probe records. No key, no account.
Base URL: http://167.71.161.149/data/tables/ (a directory listing). Dictionary as JSON: http://167.71.161.149/data/tables/tables.json. License: CC BY 4.0; upstream licenses apply per source.
DuckDB, straight off the URL, nothing to download:
INSTALL httpfs; LOAD httpfs;
SELECT route, count(*) AS requests, round(sum(billed_usd), 2) AS usd
FROM 'http://167.71.161.149/data/tables/probe_request.parquet' GROUP BY 1;
pandas:
import pandas as pd
df = pd.read_parquet('http://167.71.161.149/data/tables/market_daily.parquet')
Everything at once: download http://167.71.161.149/data/tables/routerflow.duckdb and open it with duckdb routerflow.duckdb.
A version to cite: the files here are rebuilt every hour. Monthly frozen copies live under http://167.71.161.149/data/tables/releases/<date>/, each with a MANIFEST.json of sizes and hashes, listed in http://167.71.161.149/data/tables/releases/index.json; see RELEASES.md.
A cache pair is two rows of probe_request with the same session_id: repeat = 1 (cold) and repeat = 2 (the same bytes again). What the provider recognized:
SELECT a.route, a.model, a.provider, a.endpoint, b.cached_tokens * 1.0 / b.prompt_tokens AS recognized, b.billed_usd / a.billed_usd AS second_over_first
FROM 'http://167.71.161.149/data/tables/probe_request.parquet' a JOIN 'http://167.71.161.149/data/tables/probe_request.parquet' b USING (session_id)
WHERE a.workload = 'cache-pair' AND a.repeat = 1 AND b.repeat = 2 AND NOT b.failed;
One provider across routers: provider is the OpenRouter provider slug in every table, so what a provider lists, what others measured and what we measured join on it:
SELECT q.provider, q.endpoint, q.prompt_usd_per_m AS listed, s.effective_prompt_usd_per_m AS effective, s.cache_hit_rate_1m, t.ttft_p50_ms AS trustedrouter_ttft, median(p.ttft_ms) AS our_ttft
FROM 'http://167.71.161.149/data/tables/openrouter_endpoint_quote.parquet' q
LEFT JOIN 'http://167.71.161.149/data/tables/openrouter_endpoint_stats.parquet' s ON s.snapshot = q.snapshot AND s.model = q.model AND s.endpoint = q.endpoint
LEFT JOIN 'http://167.71.161.149/data/tables/trustedrouter_provider.parquet' t ON t.snapshot = q.snapshot AND t.provider = q.provider
LEFT JOIN 'http://167.71.161.149/data/tables/probe_request.parquet' p ON p.endpoint = q.endpoint AND p.model = q.model AND NOT p.failed
WHERE q.model = 'z-ai/glm-5.3' AND q.snapshot = (SELECT max(snapshot) FROM 'http://167.71.161.149/data/tables/openrouter_endpoint_quote.parquet') GROUP BY ALL ORDER BY listed;
Every public model leaderboard we read, in one long table: usage shares from routers, arena ratings, benchmark scores, list prices, measured speeds, uptime. One row per board, period, entity and metric (board_entry), with source and board saying who publishes it under which license, board_metric giving each metric's unit and direction, and the entity tables (model, lab, provider, app, ...) and alias tables giving one id per thing. Public rows only: a board is here when its redistribution is yes or attribution; unclear and no boards are collected into a private copy that is not served, and board lists them with public = false.
board_entry is one Parquet file per source and year, board_entry/source=<source_id>/year=<yyyy>/part.parquet, listed in http://167.71.161.149/data/tables/board_entry_files.json (path, rows, bytes, period range). Over HTTP DuckDB cannot glob a directory, so read the list:
import duckdb, json, urllib.request
base = 'http://167.71.161.149/data/tables'
files = [base + '/' + f['path'] for f in json.load(urllib.request.urlopen(base + '/board_entry_files.json'))['files']]
con = duckdb.connect(); con.execute('INSTALL httpfs; LOAD httpfs')
con.read_parquet(files).create_view('board_entry')
con.execute("SELECT source_id, count(*) FROM board_entry GROUP BY 1 ORDER BY 2 DESC").fetchall()
Or download http://167.71.161.149/data/tables/routerflow.duckdb: it holds the public board_entry as a table and the views board_latest, board_current and model_everywhere. The smaller files: board_entry_latest.csv.gz (what every board shows now), and every dimension, entity and alias table as <table>.csv.gz next to its Parquet.
Snapshots. Most sources publish the same history again every night, so board_entry keeps one row per key for each run of equal values: snapshot (= first_seen) is the day a value was first collected and last_seen the newest day it was collected unchanged. A key gets a new row only when its value, interval, rank, count or cost changes; board_latest is the newest row per key. Vercel revises its past days: its share for GLM 5.3 Flash on 2026-09-19 was read as 3.76 on the 19th (the day still running, last_seen 2026-09-19) and as 4.06 from the 20th on (a second row, snapshot 2026-09-20), so a query for what Vercel says now uses board_latest, and one for what it said on a night filters snapshot <= night AND last_seen >= night.
One model across every board (model_everywhere: one row per mapped model, board and metric at its newest period):
SELECT board_id, metric, rank, value, unit, period_end, source, tier
FROM model_everywhere WHERE model_id = 'z-ai/glm-5.3-flash' ORDER BY source_id, board_id, metric;
Usage share against an arena rating over time (a router's daily share joined to the arena board that was current on each day):
WITH usage AS (SELECT period_end AS day, value AS share_pct FROM board_latest
WHERE board_id = 'vercel.leaderboard.models.day' AND metric = 'tokens' AND entity_id = 'z-ai/glm-5.3-flash'),
arena AS (SELECT period_end AS published, value AS rating, rank FROM board_latest
WHERE board_id = 'lmarena.text.overall' AND metric = 'rating' AND entity_id = 'z-ai/glm-5.3-flash')
SELECT u.day, u.share_pct, a.published AS arena_board_of, a.rating, a.rank
FROM usage u ASOF LEFT JOIN arena a ON a.published <= u.day ORDER BY u.day DESC;
The same benchmark as run by three sites (AIME: Epoch AI's own runs, MathArena and Vals; only Epoch's rows are public today, so on the served files this returns the Epoch column alone, and all three on a full build):
WITH aime AS (SELECT * FROM board_entry WHERE entity_type = 'model' AND entity_id IS NOT NULL AND lower(board_id) LIKE '%aime%' AND unit = 'pct'),
latest AS (SELECT * FROM aime QUALIFY row_number() OVER (PARTITION BY board_id, entity_id ORDER BY period_end DESC NULLS LAST, last_seen DESC, value DESC) = 1)
SELECT entity_id AS model, source_id, board_id, effort, round(value, 1) AS accuracy_pct, period_end FROM latest
WHERE entity_id IN (SELECT entity_id FROM latest GROUP BY 1 HAVING count(DISTINCT source_id) >= 3) ORDER BY model, source_id;
Names: entity_id is the canonical id (the OpenRouter slug for models) resolved from data/curated/aliases/ when the tables are built, so a corrected alias reaches all history; entity_raw is what the source printed and is never edited. name_map shows how every name resolved, or why it did not. Thinking mode, agent scaffold and free or batch tiers are columns of their own (effort, agent, variant), never part of the id.
tier column below, and tier in tables.json). Market tables are what routers and providers publish about themselves. Columns ending in _est are our estimates (dollars = tokens times a blended effective price). probe_request and quote_change are measured by us. trustedrouter_provider and attested_registry are measured by others.snapshot is the day we collected a row; date is the day the row is about. Tables keyed by snapshot grow by one copy a day, so filter to the latest snapshot unless you want the history. In tables keyed by date, snapshot says which collection the value came from: snapshot = date means the day was still running when we read it.provider is the OpenRouter provider slug and model the OpenRouter model id in every table. What a source wrote is kept beside it (provider_name, provider_raw, model_raw, provider_requested). Vercel's model display names and most of its model ids are not mapped yet.pct. Prices are dollars per million tokens.tables_schema.py, not inferred, and every build is checked against them: key unique, no table smaller than the day before. Changes are listed in CHANGELOG-DATA.md; within a major schema version they are additive, and a column that is replaced stays for one version, marked deprecated below.Generated 2026-09-22T17:28Z. Schema 1.2.1. Snapshots so far: 2026-09-18, 2026-09-19, 2026-09-20, 2026-09-21, 2026-09-22.
| table | rows | covers | what |
|---|---|---|---|
market_daily | 93 | 2026-06-21 to 2026-09-21 | OpenRouter, one row per complete UTC day: tokens served, estimated dollars, concentration (HHI, top 3) and the attested share, by tokens and by estimated dollars. |
provider_daily | 1,402 | 2026-06-21 to 2026-09-21 | OpenRouter, one row per day and provider: share of tokens and of estimated dollars. |
provider_snapshot | 440 | 2026-09-18 to 2026-09-22 | OpenRouter providers as listed on each collection day: tokens a day and a month, models, attested flag, and TrustedRouter's measurements of the same provider (tr_*). |
model_provider_window | 2,915 | 2026-09-20 to 2026-09-22 | OpenRouter, tokens a day for each model on each provider over 1, 7 and 30 complete days (the top 60 models of each window). |
openrouter_endpoint_quote | 1,172 | 2026-09-18 to 2026-09-22 | Listed prices per endpoint of OpenRouter's top 20 models on each collection day: prompt, completion, cache read, discount, quantization. |
openrouter_model_week | 100 | 2026-09-18 to 2026-09-22 | OpenRouter's weekly top 20 models: tokens, requests, share, open weights. |
openrouter_app_week | 100 | 2026-09-18 to 2026-09-22 | OpenRouter's weekly top 20 apps: tokens and requests. |
vercel_share_daily | 2,183 | 2026-07-20 to 2026-09-22 | Vercel AI Gateway, daily share of tokens by model lab. |
vercel_leaderboard_daily | 14,219 | 2026-07-21 to 2026-09-22 | Vercel AI Gateway's leaderboard export in full: daily share per model and per lab, for tokens, requests, spend, images and videos. |
vercel_endpoint | 1,708 | 2026-09-20 to 2026-09-22 | Vercel AI Gateway, every language model on every provider on each collection day: listed prices, uptime, latency, throughput. |
vercel_rank | 80 | 2026-09-20 to 2026-09-22 | Vercel AI Gateway's two top-10 rankings of apps and providers, by tokens and by spend. |
price_compare | 333 | 2026-09-20 to 2026-09-22 | The same model on the same provider, quoted by OpenRouter and by Vercel: listed input price per million tokens. |
trustedrouter_provider | 337 | 2026-09-18 to 2026-09-22 | TrustedRouter's own measurements of providers on each collection day: first-token time, availability, completion, samples. |
attested_registry | 85 | 2026-09-18 to 2026-09-22 | The awesome-private-inference registry's attestation checks per endpoint: status and layers proven. |
chutes_daily | 16 | 2026-09-07 to 2026-09-22 | Chutes, by its own meter: dollars metered a day and the share on TEE chutes. |
openrouter_endpoint_stats | 6,214 | 2026-09-19 to 2026-09-22 | Every endpoint of every OpenRouter catalog model (about 450 models, 1,500 endpoints) on each collection day: listed prices as in openrouter_endpoint_quote, plus OpenRouter's one-month effective prices, cache-hit rate and tokens per endpoint. |
openrouter_endpoint_daily | 18,778 | 2026-09-12 to 2026-09-22 | OpenRouter's daily figures per endpoint for every catalog model: cache-hit rate, median throughput, median latency, uptime. |
openrouter_model_daily | 12,515 | 2026-08-20 to 2026-09-22 | OpenRouter's daily activity per catalog model and variant: prompt, completion, cached and reasoning tokens, requests, tool calls. |
openrouter_model_app | 7,023 | 2026-09-19 to 2026-09-22 | OpenRouter's top apps for each catalog model on each collection day (up to five per model): the app by model matrix. |
probe_request | 11,799 | 2026-09-19 to 2026-09-22 | MEASURED BY US. |
quote_change | 1,245 | MEASURED BY US. | |
board_entry | 2,852,693 | 2023-02-24 to 2026-09-30 | Every leaderboard we read, in one long table: one row per board, period, entity and metric, for each value it has had. |
source | 40 | Every leaderboard source we read, public or not: who runs it, the provenance tier of its rows, its license, the citation it asks for, whether its rows may be redistributed, how and how often we read it. | |
board | 412 | Every board of every source, public or not: what it ranks, its period kind, depth, what it measures in one sentence, and the license that applies to it. | |
board_metric | 833 | One row per board and metric: the unit of board_entry.value for that metric and which direction is better. | |
board_coverage | 405 | One row per board, public or not: rows held, rows whose entity is mapped to a canonical id, first and last period, first and last snapshot, last collection time. | |
name_map | 26,157 | How every published name was resolved: one row per source, entity type, raw name, provider and join key, with the canonical ids boards/names.py gives it. | |
app | 162 | Canonical apps: OpenRouter's app slug where it has one, else the title as a slug. | |
app_alias | 172 | Every spelling of an app that a source uses. | |
family | 206 | Model families (meta/llama-3.1): the target of family aliases, for boards that name a family and not one model (Ollama pulls a library entry, not a checkpoint). | |
family_alias | 206 | Every spelling of a family that a source uses, from data/curated/aliases/family/. | |
gateway | 27 | Canonical gateways and routers that sit in front of providers (OpenRouter, Vercel AI Gateway, Cloudflare AI Gateway): the entities of gateway benchmark boards such as ComputeSDK. | |
gateway_alias | 15 | Every spelling of a gateway that a source uses, from data/curated/aliases/gateway/. | |
hardware | 21 | Canonical accelerators and systems (nvidia-h200, nvidia-b300): the entities of model by hardware boards (MLPerf) and the target of board_entry.hardware_id. | |
hardware_alias | 31 | Every spelling of a hardware that a source uses, from data/curated/aliases/hardware/. | |
lab | 430 | Canonical labs: the part of the OpenRouter slug before the slash. | |
lab_alias | 87 | Every spelling of a lab that a source uses. | |
model | 1,896 | Canonical models. | |
model_alias | 9,714 | Every spelling of a model that a source uses, and the canonical model it maps to. | |
provider | 188 | Canonical inference providers: OpenRouter's provider slug. | |
provider_alias | 700 | Every spelling of a provider that a source uses. | |
provider_model_alias | 9,264 | A provider's native model id mapped to the canonical model: the crosswalk that resolves price and speed boards, where a row is a model on a provider. | |
task | 29 | Canonical task categories (agent:memory_extraction, code-debugging): the entities of task boards such as OpenRouter's task classifications. |
market_dailyOpenRouter, one row per complete UTC day: tokens served, estimated dollars, concentration (HHI, top 3) and the attested share, by tokens and by estimated dollars.
93 rows, date 2026-06-21 to 2026-09-21. Key: date, router.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day the row is about. The collection day itself is partial and left out. [key] |
router | varchar | id | reported | Always openrouter. Nothing is summed across routers. [key] |
tokens | double | tokens | reported | Tokens served that day, summed over every provider's 90-day token chart. |
usd_est | double | usd | derived | Estimated dollars: tokens times the blended one-month effective price of each model at each provider. |
hhi_tokens | double | index | derived | Herfindahl index of provider shares of tokens, 0 to 10,000. |
hhi_usd_est | double | index | derived | Herfindahl index of provider shares of estimated dollars, 0 to 10,000. |
top3_tokens | double | fraction | derived | Share of tokens served by the three largest providers. |
top3_usd_est | double | fraction | derived | Share of estimated dollars earned by the three largest providers. |
attested_share_tokens | double | fraction | derived | Share of tokens served by attested providers. |
attested_share_usd_est | double | fraction | derived | Share of estimated dollars earned by attested providers. |
usd_priced_token_share | double | fraction | derived | Share of the day's tokens that had a model attached and so a price of their own; the rest is priced at the provider's mean. |
provider_dailyOpenRouter, one row per day and provider: share of tokens and of estimated dollars. Covers the largest providers and the attested ones; the rest are inside the daily total only.
1,402 rows, date 2026-06-21 to 2026-09-21. Key: date, router, provider.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day the row is about. [key] |
router | varchar | id | reported | Always openrouter. [key] |
provider | varchar | id | reported | Provider, as the OpenRouter provider slug (the one spelling every table uses). [key; joins provider_snapshot.provider] |
share_tokens | double | fraction | derived | Provider's share of the day's tokens. |
share_usd_est | double | fraction | derived | Provider's share of the day's estimated dollars. |
provider_snapshotOpenRouter providers as listed on each collection day: tokens a day and a month, models, attested flag, and TrustedRouter's measurements of the same provider (tr_*). The closest thing to a provider dimension until the entity tables exist.
440 rows, snapshot 2026-09-18 to 2026-09-22. Key: snapshot, router, provider.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
router | varchar | id | reported | Always openrouter. [key] |
provider | varchar | id | reported | Provider, as the OpenRouter provider slug (the one spelling every table uses). [key] |
provider_name | varchar | text | reported | Provider display name, as OpenRouter writes it. [not null] |
name | varchar | text | reported | Deprecated, use provider_name. Provider display name. |
attested | boolean | flag | derived | True when the provider is one of the attested (TEE) providers: Phala, Venice, Chutes, NEAR AI, Tinfoil, RedPill. |
models | bigint | count | reported | Models the provider serves on OpenRouter. |
tokens_day | bigint | tokens | reported | Tokens over the rolling day, from the providers page. |
tokens_30d | bigint | tokens | reported | Tokens over the rolling 30 days, from the providers page. |
share_day | double | fraction | derived | Provider's share of all providers' rolling-day tokens. |
tr_ttft_p50_ms | double | ms | measured by others | TrustedRouter's median time to first token for this provider. |
tr_availability | double | fraction | measured by others | TrustedRouter's availability for this provider. |
tr_samples | bigint | count | measured by others | Requests behind TrustedRouter's figures. |
model_provider_windowOpenRouter, tokens a day for each model on each provider over 1, 7 and 30 complete days (the top 60 models of each window).
2,915 rows, snapshot 2026-09-20 to 2026-09-22. Key: snapshot, router, period_days, model, provider.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
router | varchar | id | reported | Always openrouter. [key] |
period_days | bigint | days | reported | Length of the window: 1, 7 or 30 complete UTC days. [key] |
period_start | date | date | reported | First UTC day of the window, inclusive. |
period_end | date | date | reported | Last UTC day of the window, inclusive. |
window | varchar | text | reported | Deprecated, use period_days. The window as 1d, 7d or 30d. A reserved word in SQL. |
first | date | date | reported | Deprecated, use period_start. First day of the window. |
last | date | date | reported | Deprecated, use period_end. Last day of the window. |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). [key] |
lab | varchar | text | derived | Lab that built the model, display name, normalized by our lab map. |
provider | varchar | id | reported | Provider, as the OpenRouter provider slug (the one spelling every table uses). [key; joins provider_snapshot.provider] |
provider_name | varchar | text | reported | Provider display name. |
attested | boolean | flag | derived | True when the provider is one of the attested (TEE) providers: Phala, Venice, Chutes, NEAR AI, Tinfoil, RedPill. |
tokens_day | double | tokens_per_day | derived | Tokens a day of this model on this provider, averaged over the window. |
share_of_model | double | fraction | derived | Provider's share of the model's tokens over the window. |
openrouter_endpoint_quoteListed prices per endpoint of OpenRouter's top 20 models on each collection day: prompt, completion, cache read, discount, quantization. Paid endpoints of the standard variant, read from the model pages. For every catalog model see openrouter_endpoint_stats.
1,172 rows, snapshot 2026-09-18 to 2026-09-22. Key: snapshot, model, endpoint.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). [key] |
lab | varchar | text | derived | Lab that built the model, display name, normalized by our lab map. |
provider | varchar | id | reported | Provider, as the OpenRouter provider slug (the one spelling every table uses). Through schema 1.0 this column held the display name, which is now provider_name. [joins provider_snapshot.provider] |
provider_name | varchar | text | reported | Provider display name, as OpenRouter writes it. |
endpoint | varchar | id | reported | OpenRouter endpoint tag: the provider slug, then a quantization, region or service tier where the provider has several (deepinfra/fp8). [key] |
prompt_usd_per_m | double | usd_per_m | reported | Listed price of a million prompt tokens. |
completion_usd_per_m | double | usd_per_m | reported | Listed price of a million completion tokens. |
cache_read_usd_per_m | double | usd_per_m | reported | Listed price of a million cached prompt tokens; null where the endpoint lists none. |
discount | double | fraction | reported | Discount against list that the provider declares. |
quantization | varchar | text | reported | Quantization the endpoint declares (fp8, fp4, bf16, unknown). |
deranked | boolean | flag | reported | OpenRouter's flag for an endpoint it routes to last. |
attested | boolean | flag | derived | True when the provider is one of the attested (TEE) providers: Phala, Venice, Chutes, NEAR AI, Tinfoil, RedPill. |
openrouter_model_weekOpenRouter's weekly top 20 models: tokens, requests, share, open weights.
100 rows, snapshot 2026-09-18 to 2026-09-22. Key: snapshot, model.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
period_start | date | date | reported | First UTC day of the week, inclusive. |
period_end | date | date | reported | Last UTC day of the week, inclusive. |
window | varchar | text | reported | Deprecated, use period_end. The week as prose (week ending 2026-09-20). A reserved word in SQL. |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). [key] |
lab | varchar | text | derived | Lab that built the model, display name, normalized by our lab map. |
open_weights | boolean | flag | derived | True when the catalog links a Hugging Face repository for the model. |
tokens_week | bigint | tokens | reported | Prompt plus completion tokens over the week. |
requests_week | bigint | requests | reported | Requests over the week. |
share | double | fraction | derived | Model's share of the top 20's tokens. |
openrouter_app_weekOpenRouter's weekly top 20 apps: tokens and requests.
100 rows, snapshot 2026-09-18 to 2026-09-22. Key: snapshot, app.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
period_start | date | date | reported | First UTC day of the week, inclusive. |
period_end | date | date | reported | Last UTC day of the week, inclusive. |
window | varchar | text | reported | Deprecated, use period_end. The week as prose. A reserved word in SQL. |
app | varchar | text | reported | App title, as OpenRouter writes it. [key] |
categories | varchar[] | list | reported | OpenRouter's category tags for the app. Through schema 1.0 a comma-joined string. |
tokens_week | bigint | tokens | reported | Tokens over the week. |
requests_week | bigint | requests | reported | Requests over the week. |
vercel_share_dailyVercel AI Gateway, daily share of tokens by model lab. Shares only: Vercel publishes no totals. A subset of vercel_leaderboard_daily (lab, tokens) with lab display names, kept as it was for existing queries.
2,183 rows, date 2026-07-20 to 2026-09-22. Key: date, router, level, metric, name.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day the share is about. [key] |
router | varchar | id | reported | Always vercel. [key] |
level | varchar | id | reported | Always lab. [key] |
metric | varchar | id | reported | Always tokens. [key] |
name | varchar | text | reported | Lab that built the model, display name, normalized by our lab map. [key] |
share | double | fraction | reported | Lab's share of the gateway's tokens that day. |
snapshot | date | date | reported | UTC day of the collection the value was read from (the newest that holds the date). snapshot = date means the day was still running. [not null] |
vercel_leaderboard_dailyVercel AI Gateway's leaderboard export in full: daily share per model and per lab, for tokens, requests, spend, images and videos. Upstream keeps a rolling 61 days; this table keeps every day ever collected. Models are the ones Vercel names (about ten a day per metric); the rest are inside the row named Other.
14,219 rows, date 2026-07-21 to 2026-09-22. Key: date, level, metric, name. Keeps every day ever collected.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day the share is about. [key] |
router | varchar | id | reported | Always vercel. |
level | varchar | id | reported | model or lab. [key] |
metric | varchar | id | reported | tokens, requests, spend, images or videos (Vercel writes imageCount and videoCount; see metric_raw). [key] |
metric_raw | varchar | id | reported | Metric as Vercel writes it. [not null] |
name | varchar | text | reported | Model display name or lab slug, exactly as Vercel writes it (GLM 5.3 Flash; zai). [key] |
lab | varchar | text | derived | For lab rows, the lab display name by our lab map (zai becomes Z.ai). Null for model rows: mapping Vercel's model names waits for the alias tables. |
residual | boolean | flag | derived | True for the row named Other: everything Vercel did not name in that export. The named set moves as the 61-day window rolls, so after the union the named rows of a day can add up to more than 100 minus Other; 1 - sum(share of the named rows) is always exact. [not null] |
share | double | fraction | reported | Share of the gateway's total for that metric and day (share_percent / 100). [not null] |
snapshot | date | date | reported | UTC day of the collection this value was read from: the newest one that published it. A value with snapshot = date was read while that day was still running and is partial; it is replaced the next night. |
vercel_endpointVercel AI Gateway, every language model on every provider on each collection day: listed prices, uptime, latency, throughput. Through schema 1.0 it held the newest day only; it now keeps every day from 2026-09-20 on.
1,708 rows, snapshot 2026-09-20 to 2026-09-22. Key: snapshot, model_raw, provider_raw. Keeps every day ever collected.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
router | varchar | id | reported | Always vercel. |
model | varchar | id | derived | Model, as the OpenRouter model id (the one spelling every table uses). Only the lab prefix zai/ is rewritten (to z-ai/); other Vercel ids are as published until the alias tables exist. |
model_raw | varchar | id | reported | Model id exactly as Vercel writes it (zai/glm-5.3). [key] |
provider | varchar | id | derived | Provider, as the OpenRouter provider slug (the one spelling every table uses). Vercel's bedrock, vertex, google, zai, togetherai and moonshot are mapped; the rest are as published. [joins provider_snapshot.provider] |
provider_raw | varchar | id | reported | Provider exactly as Vercel writes it (bedrock). [key] |
prompt_usd_per_m | double | usd_per_m | reported | Listed price of a million prompt tokens. |
completion_usd_per_m | double | usd_per_m | reported | Listed price of a million completion tokens. |
cache_read_usd_per_m | double | usd_per_m | reported | Listed price of a million cached prompt tokens. |
quantization | varchar | text | reported | Quantization, where Vercel states one (so far it never does). |
zdr | boolean | flag | reported | Vercel lists the endpoint as zero data retention. |
implicit_caching | boolean | flag | reported | Vercel lists the endpoint as caching prompts without being asked. |
uptime_1d | double | pct | reported | Uptime over the last day, 0 to 100, as Vercel reports it. |
latency_p50_ms | double | ms | reported | Median latency over the last hour, as Vercel reports it. |
throughput_p50 | double | tokens_per_s | reported | Median throughput over the last hour, as Vercel reports it. |
vercel_rankVercel AI Gateway's two top-10 rankings of apps and providers, by tokens and by spend. Ranks only.
80 rows, snapshot 2026-09-20 to 2026-09-22. Key: snapshot, kind, name.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
kind | varchar | id | reported | app or provider. [key] |
name | varchar | text | reported | App or provider display name, as Vercel writes it. [key] |
rank_by_tokens | bigint | rank | reported | Rank by token volume; null when outside that top 10. |
rank_by_spend | bigint | rank | reported | Rank by spend; null when outside that top 10. |
price_compareThe same model on the same provider, quoted by OpenRouter and by Vercel: listed input price per million tokens.
333 rows, snapshot 2026-09-20 to 2026-09-22. Key: snapshot, model, provider.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | derived | UTC day we collected the row. [key] |
model | varchar | id | derived | Model, as the OpenRouter model id (the one spelling every table uses). [key] |
provider | varchar | id | derived | Provider, as the OpenRouter provider slug (the one spelling every table uses). [key; joins provider_snapshot.provider] |
provider_name | varchar | text | derived | Provider display name, as OpenRouter writes it. |
openrouter_endpoint | varchar | id | derived | The OpenRouter endpoint compared: the provider's plain one where it has one, else its cheapest variant. [joins openrouter_endpoint_quote.endpoint] |
openrouter_prompt_usd_per_m | double | usd_per_m | reported | OpenRouter's listed prompt price. |
vercel_prompt_usd_per_m | double | usd_per_m | reported | Vercel's listed prompt price. |
vercel_vs_openrouter | double | fraction | derived | Vercel's price over OpenRouter's, minus one. |
openrouter_cache_usd_per_m | double | usd_per_m | reported | OpenRouter's listed cache-read price. |
vercel_cache_usd_per_m | double | usd_per_m | reported | Vercel's listed cache-read price. |
trustedrouter_providerTrustedRouter's own measurements of providers on each collection day: first-token time, availability, completion, samples. Measured by TrustedRouter, not by us.
337 rows, snapshot 2026-09-18 to 2026-09-22. Key: snapshot, provider_raw.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | measured by others | UTC day we collected the row. [key] |
provider | varchar | id | derived | Provider, as the OpenRouter provider slug (the one spelling every table uses). Through schema 1.0 this column held TrustedRouter's own slug, which is now provider_raw; the two differ for the nine providers of the TR_SLUG map in build_dataset.py (zai, gmi, kimi, grok, nvidia-nim and others). [joins provider_snapshot.provider] |
provider_raw | varchar | id | measured by others | Provider slug exactly as TrustedRouter writes it. [key] |
rank | bigint | rank | measured by others | TrustedRouter's rank; null for providers it does not qualify. |
ttft_p50_ms | double | ms | measured by others | Median time to first token. |
ttft_p95_ms | double | ms | measured by others | 95th percentile time to first token. |
throughput_tps | double | tokens_per_s | measured by others | Throughput, as TrustedRouter measures it. |
completion | double | fraction | measured by others | Share of requests that completed. |
availability | double | fraction | measured by others | Share of requests the provider accepted. |
samples | bigint | count | measured by others | Requests behind the figures. |
models | bigint | count | measured by others | Models measured. |
verified_confidential | boolean | flag | measured by others | TrustedRouter labels the provider as verified confidential. |
zdr_label | boolean | flag | measured by others | TrustedRouter labels the provider as zero data retention. |
attested_registryThe awesome-private-inference registry's attestation checks per endpoint: status and layers proven.
85 rows, snapshot 2026-09-18 to 2026-09-22. Key: snapshot, provider, endpoint.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | measured by others | UTC day we collected the row. [key] |
provider | varchar | id | measured by others | Provider, as the registry writes it; its ids match OpenRouter's slugs where the provider is on OpenRouter. [key] |
endpoint | varchar | id | measured by others | The model or endpoint name the registry checks. [key] |
status | varchar | text | measured by others | Registry verdict: verified, partial, invalid. |
layers_proven | bigint | count | measured by others | Attestation layers the registry could prove. |
layers_required | bigint | count | measured by others | Layers it requires. |
missing | varchar | text | measured by others | Layers not proven, comma-joined. |
chutes_dailyChutes, by its own meter: dollars metered a day and the share on TEE chutes.
16 rows, date 2026-09-07 to 2026-09-22. Key: date.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day. The newest day is partial. [key] |
usd_metered | double | usd | reported | Dollars Chutes metered that day. |
share_tee | double | fraction | derived | Share of those dollars on chutes whose name marks them as TEE. |
share_plain | double | fraction | derived | Share on chutes listed without TEE. |
share_unlisted | double | fraction | derived | Share on chutes missing from the public list. |
openrouter_endpoint_statsEvery endpoint of every OpenRouter catalog model (about 450 models, 1,500 endpoints) on each collection day: listed prices as in openrouter_endpoint_quote, plus OpenRouter's one-month effective prices, cache-hit rate and tokens per endpoint. From the frontend stats behind the model pages. Model ids that start with ~ are OpenRouter's "latest" aliases and repeat another model's endpoints.
6,214 rows, snapshot 2026-09-19 to 2026-09-22. Key: snapshot, model_permaslug, variant, endpoint_id. Keeps every day ever collected.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). With the variant as a suffix where it is not standard (:free, :batch). [not null] |
model_permaslug | varchar | id | reported | OpenRouter's dated permanent slug for the model. [key] |
variant | varchar | id | reported | standard, free or batch. [key] |
lab | varchar | text | derived | Lab that built the model, display name, normalized by our lab map. |
provider | varchar | id | reported | Provider, as the OpenRouter provider slug (the one spelling every table uses). [joins provider_snapshot.provider] |
provider_name | varchar | text | reported | Provider display name. |
endpoint | varchar | id | reported | OpenRouter endpoint tag (deepinfra/fp8). |
endpoint_id | varchar | id | reported | OpenRouter's id for the endpoint; the key of openrouter_endpoint_daily. [key] |
prompt_usd_per_m | double | usd_per_m | reported | Listed price of a million prompt tokens. |
completion_usd_per_m | double | usd_per_m | reported | Listed price of a million completion tokens. |
cache_read_usd_per_m | double | usd_per_m | reported | Listed price of a million cached prompt tokens. |
discount | double | fraction | reported | Discount against list that the provider declares. |
quantization | varchar | text | reported | Quantization the endpoint declares. |
deranked | boolean | flag | reported | OpenRouter's flag for an endpoint it routes to last. |
attested | boolean | flag | derived | True when the provider is one of the attested (TEE) providers: Phala, Venice, Chutes, NEAR AI, Tinfoil, RedPill. |
is_free | boolean | flag | reported | A free endpoint. |
is_disabled | boolean | flag | reported | OpenRouter lists the endpoint as disabled. |
top20 | boolean | flag | derived | The model was in OpenRouter's weekly top 20 that day. [not null] |
effective_prompt_usd_per_m | double | usd_per_m | reported | OpenRouter's effective price of a million prompt tokens over the last month, after cache hits and discounts. |
effective_completion_usd_per_m | double | usd_per_m | reported | Effective price of a million completion tokens over the last month. |
cache_hit_rate_1m | double | fraction | reported | Cache-hit rate over the last month, as OpenRouter reports it. |
tokens_1m | bigint | tokens | reported | Tokens the endpoint served over the last month. |
openrouter_endpoint_dailyOpenRouter's daily figures per endpoint for every catalog model: cache-hit rate, median throughput, median latency, uptime. Upstream keeps about 8 days (3 for uptime); this table keeps every day ever collected.
18,778 rows, date 2026-09-12 to 2026-09-22. Key: date, model_permaslug, variant, endpoint_id. Keeps every day ever collected.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day the figures are about. [key] |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). [not null] |
model_permaslug | varchar | id | reported | OpenRouter's dated permanent slug. [key] |
variant | varchar | id | reported | standard, free or batch. [key] |
endpoint_id | varchar | id | reported | OpenRouter's id for the endpoint. The literal openrouter marks OpenRouter's own uptime for the model across all its endpoints. [key; joins openrouter_endpoint_stats.endpoint_id] |
provider | varchar | id | reported | Provider, as the OpenRouter provider slug (the one spelling every table uses). Null on the endpoint_id = openrouter rows, and for an endpoint that had left every endpoint list by the time we collected. [joins provider_snapshot.provider] |
endpoint | varchar | id | reported | OpenRouter endpoint tag (deepinfra/fp8). |
cache_hit_rate | double | fraction | reported | Share of prompt tokens served from cache that day, as OpenRouter reports it (published in percent). |
throughput_p50_tps | double | tokens_per_s | reported | Median throughput that day. |
latency_p50_ms | double | ms | reported | Median latency to first token that day. |
uptime | double | fraction | reported | Uptime that day (published in percent). |
snapshot | date | date | reported | UTC day of the collection this value was read from: the newest one that published it. A value with snapshot = date was read while that day was still running and is partial; it is replaced the next night. |
openrouter_model_dailyOpenRouter's daily activity per catalog model and variant: prompt, completion, cached and reasoning tokens, requests, tool calls. Upstream keeps 31 days; this table keeps every day ever collected.
12,515 rows, date 2026-08-20 to 2026-09-22. Key: date, model_permaslug, variant. Keeps every day ever collected.
| column | type | unit | tier | description |
|---|---|---|---|---|
date | date | date | reported | UTC day the figures are about. [key] |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). [not null] |
model_permaslug | varchar | id | reported | OpenRouter's dated permanent slug. [key] |
variant | varchar | id | reported | standard, free or batch. [key] |
lab | varchar | text | derived | Lab that built the model, display name, normalized by our lab map. |
prompt_tokens | bigint | tokens | reported | Prompt tokens that day. |
completion_tokens | bigint | tokens | reported | Completion tokens that day. |
cached_tokens | bigint | tokens | reported | Prompt tokens served from cache (native count). |
reasoning_tokens | bigint | tokens | reported | Reasoning tokens (native count). |
requests | bigint | requests | reported | Requests that day. |
tool_calls | bigint | count | reported | Tool calls that day. |
tool_call_error_requests | bigint | requests | reported | Requests with a tool-call error. |
media_prompts | bigint | count | reported | Images and other media sent in prompts. |
snapshot | date | date | reported | UTC day of the collection this value was read from: the newest one that published it. A value with snapshot = date was read while that day was still running and is partial; it is replaced the next night. |
openrouter_model_appOpenRouter's top apps for each catalog model on each collection day (up to five per model): the app by model matrix. The window is undocumented upstream and longer than a week.
7,023 rows, snapshot 2026-09-19 to 2026-09-22. Key: snapshot, model_permaslug, variant, app_id. Keeps every day ever collected.
| column | type | unit | tier | description |
|---|---|---|---|---|
snapshot | date | date | reported | UTC day we collected the row. [key] |
model | varchar | id | reported | Model, as the OpenRouter model id (the one spelling every table uses). [not null] |
model_permaslug | varchar | id | reported | OpenRouter's dated permanent slug. [key] |
variant | varchar | id | reported | standard, free or batch. [key] |
app_id | bigint | id | reported | OpenRouter's id for the app. [key] |
app | varchar | text | reported | App title, as OpenRouter writes it. |
categories | varchar[] | list | reported | OpenRouter's category tags for the app. |
tokens | bigint | tokens | reported | Tokens of this model used by the app over OpenRouter's window. |
requests | bigint | requests | reported | Requests over the same window. |
probe_requestMEASURED BY US. One row per request the prober sent, through OpenRouter, TrustedRouter or Vercel (route): tokens, cached tokens, the bill, first-token time, who served it. No prompt or response text. A cache pair is two rows with the same session_id, repeat 1 and 2. Early and unreviewed.
11,799 rows, date 2026-09-19 to 2026-09-22. Key: ts, session_id.
| column | type | unit | tier | description |
|---|---|---|---|---|
ts | timestamp | timestamp | measured by us | When the request was sent, UTC. [key] |
date | date | date | measured by us | UTC day of ts. |
route | varchar | id | measured by us | Router the request went through: openrouter, trustedrouter or vercel. |
job | varchar | id | measured by us | Prober job that sent it. |
run_id | varchar | id | measured by us | One run of a job. |
session_id | varchar | id | measured by us | One conversation; the rows of a cache pair or of an agent trace share it. [key] |
workload | varchar | id | measured by us | Workload: cache-pair, agent-trace and the rest of the prober's taxonomy. |
family | varchar | id | measured by us | Workload family. |
model | varchar | id | measured by us | Model, as the OpenRouter model id (the one spelling every table uses). |
pinned | boolean | flag | measured by us | True when the request named one provider and allowed no fallback. |
provider_requested | varchar | id | measured by us | What the prober asked the router for, in that router's own spelling: an OpenRouter endpoint tag (deepinfra/fp8), or a provider slug on the other routes. Null when unpinned. |
provider | varchar | id | derived | Provider, as the OpenRouter provider slug (the one spelling every table uses). From provider_requested; null when unpinned. [joins provider_snapshot.provider] |
endpoint | varchar | id | derived | OpenRouter endpoint tag requested (deepinfra/fp8); null on other routes and when unpinned. [joins openrouter_endpoint_stats.endpoint] |
provider_served | varchar | text | reported | Provider the router says served the request, display name as the router writes it. |
turn | bigint | count | measured by us | Turn within a multi-turn session. |
repeat | bigint | count | measured by us | 1 for the first send, 2 for the same bytes again. |
arm | varchar | id | measured by us | Experiment arm. |
request_name | varchar | id | measured by us | Name of the request within its workload. |
event_id | varchar | id | measured by us | Event the request belongs to. |
repeat_gap_s | double | s | measured by us | Seconds between the first send and the repeat. |
repeated_prompt_tokens | bigint | tokens | measured by us | Prompt tokens that repeat an earlier request in the session. |
prompt_tokens | bigint | tokens | reported | Prompt tokens, as the router's usage block reports. |
completion_tokens | bigint | tokens | reported | Completion tokens, as reported. |
cached_tokens | bigint | tokens | reported | Prompt tokens the provider recognized from cache, as reported. |
reasoning_tokens | bigint | tokens | reported | Reasoning tokens, as reported. |
billed_usd | double | usd | measured by us | What the request cost. |
bill_reported_by_router | boolean | flag | measured by us | True when the router reported the bill; false when we computed it from tokens and quotes. |
quote_prompt_usd_per_m | double | usd_per_m | reported | Listed prompt price at send time. |
quote_cache_read_usd_per_m | double | usd_per_m | reported | Listed cache-read price at send time. |
quantization | varchar | text | reported | Quantization the endpoint declared at send time. |
ttft_ms | double | ms | measured by us | Time to first token, measured at the prober. |
total_ms | double | ms | measured by us | Time to the last byte. |
tokens_per_s | double | tokens_per_s | measured by us | Completion tokens per second after the first token. |
finish_reason | varchar | text | reported | Finish reason the router returned. |
http_status | bigint | id | measured by us | HTTP status. |
failed | boolean | flag | measured by us | The request errored. |
check_passed | boolean | flag | measured by us | The answer passed the workload's check; null where there is none. |
workload_hash | varchar | id | measured by us | Hash of the workload file the request was built from: equal hashes mean the same workload version. |
response_sha256 | varchar | id | measured by us | SHA-256 of the response text. The text itself is never exported. |
quote_changeMEASURED BY US. Every change in a quoted price that the 5-minute poll of OpenRouter saw, for the panel models.
1,245 rows. Key: ts, model, endpoint, field.
| column | type | unit | tier | description |
|---|---|---|---|---|
ts | timestamp | timestamp | measured by us | When the poll saw the change, UTC. [key] |
model | varchar | id | measured by us | Model, as the OpenRouter model id (the one spelling every table uses). [key] |
provider | varchar | id | derived | Provider, as the OpenRouter provider slug (the one spelling every table uses). [not null; joins provider_snapshot.provider] |
endpoint | varchar | id | measured by us | OpenRouter endpoint tag (deepinfra/fp8). [key; joins openrouter_endpoint_stats.endpoint] |
field | varchar | id | measured by us | What changed: prompt_usd_per_m, completion_usd_per_m, cache_read_usd_per_m, or endpoint (listed or delisted). [key] |
old | double | usd_per_m | measured by us | Value before, for price fields; null for field = endpoint. Through schema 1.0 typed JSON. |
new | double | usd_per_m | measured by us | Value after, for price fields; null for field = endpoint. Through schema 1.0 typed JSON. |
old_text | varchar | text | measured by us | Value before, as text, for every field (null, listed). |
new_text | varchar | text | measured by us | Value after, as text, for every field (listed, delisted). |
board_entryEvery leaderboard we read, in one long table: one row per board, period, entity and metric, for each value it has had. Usage shares, arena ratings, benchmark scores, prices and measured speeds from about 40 sources, with history where upstream keeps it. One Parquet file per source and year (board_entry/source=<id>/year=<yyyy>/part.parquet, listed in board_entry_files.json). A source publishes here only if its license allows redistribution; the rest is collected into files that are not served. The tier of a row is its source's (source.tier).
2,852,693 rows, period_end 2023-02-24 to 2026-09-30. Key: board_id, period_end, entity_raw, provider_raw, effort, agent, variant, hardware, metric, snapshot. Keeps every day ever collected. 32 files over 12 sources, listed in board_entry_files.json; snapshots 2024-10-22 to 2026-09-22; the private copy holds 3,718,227 rows of 38 sources.
| column | type | unit | tier | description |
|---|---|---|---|---|
source_id | varchar | id | per source | Source the row came from: the first segment of board_id. Also the directory the file lives in (source=<source_id>). [not null; joins source.id] |
board_id | varchar | id | per source | Board the row belongs to, <source_id>.<name>. Part of the row key. [key; joins board.id] |
snapshot | date | date | per source | UTC day we first collected this value for this key: the snapshot that is kept. A key gets a new row only when its value, interval, rank, n or cost changes, so most sources, which publish the same history again every night, cost one row per value and not one per night. Part of the row key. [key] |
first_seen | date | date | per source | The same day as snapshot, under the name that pairs with last_seen. [not null] |
last_seen | date | date | per source | Newest UTC day on which we collected the same value for this key. Earlier than the newest snapshot of the board when the entry left the board or its value changed (the next row of the key says which). [not null] |
collected_at | timestamp | timestamp | per source | UTC time of the collection on first_seen. [not null] |
last_collected_at | timestamp | timestamp | per source | UTC time of the collection on last_seen. [not null] |
period_start | date | date | per source | First day the value covers; null on boards dated as of a day (asof) and on boards with no date of their own. |
period_end | date | date | per source | Last day the value covers, or the day upstream published it; null on boards that only show "now", which are told apart by snapshot. The file's year= directory is the year of period_end, else of snapshot. Part of the row key. [key] |
rank | integer | rank | per source | Position on the board for this metric and period, 1 is top; null for residual rows and unranked values. |
entity_type | varchar | id | per source | What is ranked: model, lab, app, provider, endpoint, router, gateway, hardware, task, language, country, system, or residual (the row named Other, kept so shares still sum to 100). [not null] |
entity_raw | varchar | text | per source | Name exactly as the source published it, never edited. Part of the row key. [key] |
entity_id | varchar | id | derived | Canonical id of the entity: joins the table named by entity_type (model.id, lab.id, provider.id, app.id, gateway.id, task.id, hardware.id). A model row that no model alias answers may resolve to a family (family.id, name_map.method = family). Null until the name is mapped. Resolved when the tables are built, from data/curated/aliases/, so a corrected alias applies to all history. [joins model.id] |
provider_raw | varchar | text | per source | Serving provider exactly as published, on boards whose row is a model on a provider (prices, speeds, uptime). Part of the row key. [key] |
provider_id | varchar | id | derived | Canonical provider id; null until mapped. [joins provider.id] |
effort | varchar | text | per source | Thinking mode or reasoning effort as published (high, thinking, 16k). Part of the row key. [key] |
agent | varchar | text | per source | Agent scaffold or harness the model ran in (SWE-agent, Terminus 2). Part of the row key. [key] |
variant | varchar | text | per source | Anything else that splits rows of one entity: free, batch, quantization, checkpoint date, scenario. Part of the row key. [key] |
variant_from_name | boolean | flag | derived | True when the source published no variant and the name mapping took one off the name (:free, a checkpoint date). [not null] |
hardware | varchar | text | per source | Accelerator or system, on model by hardware boards. Part of the row key. [key] |
hardware_id | varchar | id | derived | Canonical id of the hardware column (8x NVIDIA H200-SXM-141GB gives nvidia-h200; a system that mixes accelerators has none). Not in the row contract: derived at build time through boards/names.py, like entity_id. [joins hardware.id] |
metric | varchar | id | per source | Metric name, declared per board in board_metric. Part of the row key. [key; joins board_metric.metric] |
value | double | per_row | per source | The number, in the unit of the unit column. Percentages are 0 to 100; prices are dollars per million tokens. [not null] |
value_low | double | per_row | per source | Lower bound of the published interval. |
value_high | double | per_row | per source | Upper bound of the published interval. |
unit | varchar | id | per source | Unit of value: tokens, requests, usd, usd_per_m, share_pct, rank, rating, pct, index, runs, downloads, ms, s, minutes, tokens_per_s or count. [not null] |
n | bigint | count | per source | Votes, samples or trials behind the value. |
cost_usd | double | usd | per source | Cost per task or run in dollars, where the board reports it next to the score. |
source_url | varchar | url | per source | URL the raw bytes came from, on first_seen. Never holds a key. [not null] |
extra | json | json | per source | Small flat object of leftovers the source published (organization, license, upstream ids). partial = true marks a period that was still running when collected. |
sourceEvery leaderboard source we read, public or not: who runs it, the provenance tier of its rows, its license, the citation it asks for, whether its rows may be redistributed, how and how often we read it.
40 rows. Key: id. Also as source.csv.gz. Every source and board, public or not.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Source id, the first segment of every board_id. [key] |
name | varchar | text | derived | Name of the source. |
operator | varchar | text | derived | Who runs it. |
homepage | varchar | url | derived | Where a person would look at it. |
kind | varchar | id | derived | router, app, eval, measurement or catalog. |
tier | varchar | id | derived | Provenance tier of its rows: reported, measured_by_others, measured_by_us or derived. |
license | varchar | text | derived | License of the data, as stated upstream ("none stated" is an answer). |
license_url | varchar | url | derived | Where the license is stated. |
redistribution | varchar | id | derived | yes, attribution, unclear or no. Boards of a source marked unclear or no are not in the public board_entry; a board can override its source (board.redistribution). |
citation | varchar | text | derived | Citation line the source asks for. Required wherever redistribution is attribution. |
terms_url | varchar | url | derived | Terms of use, where they matter. |
access | varchar | id | derived | How we read it: json_api, file, git, dataset, embedded_state or html. |
auth | varchar | id | derived | none, or the name of the environment variable that holds the key. |
cadence | varchar | id | derived | How often upstream changes: hourly, daily, weekly, monthly, irregular. We look every night. |
upstream_history | varchar | id | derived | What upstream itself keeps: none, rolling or full. |
upstream_history_note | varchar | text | derived | The window, where it is rolling. |
notes | varchar | text | derived | Caveats the source module states. |
boards | integer | count | derived | Boards the source declares. [not null] |
public_boards | integer | count | derived | Of those, boards whose rows are published. [not null] |
public | boolean | flag | derived | True when at least one of its boards is published. [not null] |
declared | boolean | flag | derived | False for a source id found in the rows that no source module declares; its rows stay private. [not null] |
boardEvery board of every source, public or not: what it ranks, its period kind, depth, what it measures in one sentence, and the license that applies to it.
412 rows. Key: id. Also as board.csv.gz. Every source and board, public or not.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Board id, <source_id>.<name>. [key] |
source_id | varchar | id | derived | Source of the board. [not null; joins source.id] |
title | varchar | text | derived | Title of the board. |
url | varchar | url | derived | Where a person would look at it. |
entity_type | varchar | id | derived | What the board ranks. |
category | varchar | id | derived | Category or task the board is about (programming, vision), where it has one. |
period | varchar | id | derived | day, week, month, rolling (another trailing window), asof (a state dated by upstream) or none (no date upstream: snapshot is the only time). |
metrics | varchar[] | list | derived | Metric names of the board; units are in board_metric. |
rank_metric | varchar | id | derived | Metric the board's own ordering follows. |
higher_is_better | boolean | flag | derived | Direction of the board's metrics; a metric can override it (board_metric). |
sums_to_100 | boolean | flag | derived | The shares of one period and metric add up to 100. |
depth | varchar | text | derived | How much of the ranking is published: top N, or full. |
measures | varchar | text | derived | What the board measures, in one sentence. |
tier | varchar | id | derived | Provenance tier, from the source. |
license | varchar | text | derived | License that applies: the board's own where it has one, else the source's. |
license_url | varchar | url | derived | Where that license is stated. |
redistribution | varchar | id | derived | yes, attribution, unclear or no: the board's own where it has one, else the source's. |
citation | varchar | text | derived | Citation line that applies. |
notes | varchar | text | derived | Caveats the source module states. |
public | boolean | flag | derived | True when redistribution is yes or attribution: the board's rows are in the public board_entry. [not null] |
declared | boolean | flag | derived | False for a board id found in the rows that no source module declares; its rows stay private. [not null] |
board_metricOne row per board and metric: the unit of board_entry.value for that metric and which direction is better.
833 rows. Key: board_id, metric. Also as board_metric.csv.gz. Every source and board, public or not.
| column | type | unit | tier | description |
|---|---|---|---|---|
board_id | varchar | id | derived | Board the metric belongs to. [key; joins board.id] |
metric | varchar | id | derived | Metric name, as in board_entry.metric. [key] |
unit | varchar | id | derived | Unit of the value, from the closed list of boards/contract.py. [not null] |
higher_is_better | boolean | flag | derived | The metric's own direction where declared, else the board's. |
description | varchar | text | derived | What the metric is, where the source says. |
board_coverageOne row per board, public or not: rows held, rows whose entity is mapped to a canonical id, first and last period, first and last snapshot, last collection time. Counts of private boards are metadata and are listed; their rows are not.
405 rows. Key: board_id. Also as board_coverage.csv.gz. Every source and board, public or not.
| column | type | unit | tier | description |
|---|---|---|---|---|
board_id | varchar | id | derived | Board the row counts. [key; joins board.id] |
source_id | varchar | id | derived | Source of the board. [not null; joins source.id] |
rows | bigint | count | derived | Rows of board_entry. [not null] |
mappable_rows | bigint | count | derived | Rows whose entity type has an entity table (residual rows and types without a table cannot be mapped). [not null] |
mapped_rows | bigint | count | derived | Of those, rows with an entity_id. [not null] |
entities | bigint | count | derived | Distinct entity_raw. [not null] |
mapped_entities | bigint | count | derived | Distinct entity_id. [not null] |
first_period_end | date | date | derived | Oldest period_end; null on boards with no date of their own. |
last_period_end | date | date | derived | Newest period_end. |
first_snapshot | date | date | derived | Oldest snapshot kept. [not null] |
last_snapshot | date | date | derived | Newest day the board was collected (max of last_seen). [not null] |
last_collected_at | timestamp | timestamp | derived | UTC time of that collection: the board's last sync. [not null] |
public | boolean | flag | derived | The board's rows are published. [not null] |
name_mapHow every published name was resolved: one row per source, entity type, raw name, provider and join key, with the canonical ids boards/names.py gives it. board_entry.entity_id and provider_id are this table joined to the rows at build time. Names of private boards are in the private copy only.
26,157 rows. Key: source_id, entity_type, entity_raw, provider_raw, join_key. Also as name_map.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
source_id | varchar | id | derived | Source whose spelling this is. [key; joins source.id] |
entity_type | varchar | id | derived | Entity type of the rows. [key] |
entity_raw | varchar | text | derived | Name as published. [key] |
provider_raw | varchar | text | derived | Provider as published, on model by provider rows. [key] |
join_key | varchar | json | derived | The ids the source publishes next to the name and boards/names.py joins on (from board_entry.extra), as a JSON object; null where the source has none. [key] |
entity_id | varchar | id | derived | Canonical id, null when unmapped. [joins model.id] |
provider_id | varchar | id | derived | Canonical provider id, null when unmapped or no provider. [joins provider.id] |
variant_from_name | varchar | text | derived | Variant the mapping took off the name (free, a checkpoint date). |
method | varchar | text | derived | How the entity resolved: alias, seed, crosswalk, id, name; else why not: no alias, ambiguous, id differs, no entity table (the entity type has none), proposed (the alias points at an id no entity table holds yet). |
provider_method | varchar | text | derived | The same for the provider. |
proposed_id | varchar | id | derived | The id an alias points at when no entity table holds it yet (method = proposed). |
rows | bigint | count | derived | Rows of board_entry with this name. |
appCanonical apps: OpenRouter's app slug where it has one, else the title as a slug. Apps are products, never people.
162 rows. Key: id. Also as app.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical id: lowercase, stable, never reused. [key] |
display_name | varchar | text | derived | Name for people. |
homepage | varchar | url | derived | Home page, where known. |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
app_aliasEvery spelling of an app that a source uses.
172 rows. Key: file, line. Also as app_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | App name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins app.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
openrouter_app_id | varchar | text | derived | Column of data/curated/app_alias that the schema does not declare yet; passed through as text. |
openrouter_slug | varchar | text | derived | Column of data/curated/app_alias that the schema does not declare yet; passed through as text. |
familyModel families (meta/llama-3.1): the target of family aliases, for boards that name a family and not one model (Ollama pulls a library entry, not a checkpoint). A model row that no model alias answers may carry a family id in entity_id, marked family in name_map.method.
206 rows. Key: id. Also as family.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical family id, lab/family. [key] |
display_name | varchar | text | derived | Name for people. |
lab_id | varchar | id | derived | Lab of the family. [joins lab.id] |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
family_aliasEvery spelling of a family that a source uses, from data/curated/aliases/family/.
206 rows. Key: file, line. Also as family_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | Family name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins family.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
gatewayCanonical gateways and routers that sit in front of providers (OpenRouter, Vercel AI Gateway, Cloudflare AI Gateway): the entities of gateway benchmark boards such as ComputeSDK.
27 rows. Key: id. Also as gateway.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical gateway id. [key] |
display_name | varchar | text | derived | Name for people. |
homepage | varchar | url | derived | Home page of the gateway. |
kind | varchar | id | derived | gateway or router, as the registry files it. |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
gateway_aliasEvery spelling of a gateway that a source uses, from data/curated/aliases/gateway/.
15 rows. Key: file, line. Also as gateway_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | Gateway name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins gateway.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
hardwareCanonical accelerators and systems (nvidia-h200, nvidia-b300): the entities of model by hardware boards (MLPerf) and the target of board_entry.hardware_id.
21 rows. Key: id. Also as hardware.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical hardware id. [key] |
display_name | varchar | text | derived | Name for people. |
vendor | varchar | id | derived | Vendor slug. |
kind | varchar | id | derived | accelerator or system. |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
hardware_aliasEvery spelling of a hardware that a source uses, from data/curated/aliases/hardware/.
31 rows. Key: file, line. Also as hardware_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | Hardware name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins hardware.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
labCanonical labs: the part of the OpenRouter slug before the slash.
430 rows. Key: id. Also as lab.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical id: lowercase, stable, never reused. [key] |
display_name | varchar | text | derived | Name for people. |
homepage | varchar | url | derived | Home page, where known. |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
lab_aliasEvery spelling of a lab that a source uses.
87 rows. Key: file, line. Also as lab_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | Lab name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins lab.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
modelCanonical models. id is the OpenRouter slug where one exists, else the lowercased Hugging Face repo id, else lab/name minted by hand. Thinking mode, effort, agent, free tier, quantization and checkpoint dates are never in the id.
1,896 rows. Key: id. Also as model.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical model id. [key] |
display_name | varchar | text | derived | Name for people. |
lab_id | varchar | id | derived | Lab that built it. [joins lab.id] |
open_weights | boolean | flag | derived | True where a catalog names a Hugging Face repository; null is unknown, not false. |
hf_repo | varchar | id | derived | Hugging Face repository. |
family | varchar | id | derived | Model family, a convenience, not a key. |
released | date | date | derived | Release day (the day OpenRouter listed it, where that is the origin). |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
modality | varchar | id | derived | What the model takes and makes: text, image, video, audio, embedding, or a combination as the catalog states it. |
model_aliasEvery spelling of a model that a source uses, and the canonical model it maps to. Curated as CSV under data/curated/aliases/model/ so a change is a reviewable diff.
9,714 rows. Key: file, line. Also as model_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | Model name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins model.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
kilo_slug | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
openrouter_id | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
aa_slug | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
openrouter_permaslug | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
da_model_id | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
vercel_id | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
origin | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
aa_id | varchar | text | derived | Column of data/curated/model_alias that the schema does not declare yet; passed through as text. |
providerCanonical inference providers: OpenRouter's provider slug. The provider column of the older tables uses the same ids.
188 rows. Key: id. Also as provider.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical id: lowercase, stable, never reused. [key] |
display_name | varchar | text | derived | Name for people. |
homepage | varchar | url | derived | Home page, where known. |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |
provider_aliasEvery spelling of a provider that a source uses.
700 rows. Key: file, line. Also as provider_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/ the row comes from. <source_id>.csv is the source's own file; _seed_<origin>.csv is a crosswalk found in raw files. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source whose spelling this is. [joins source.id] |
raw_name | varchar | text | derived | Provider name as that source writes it. |
entity_id | varchar | id | derived | Canonical id the name maps to. An id that no entity table holds yet is a proposal and is not applied to board_entry. [joins provider.id] |
variant | varchar | text | derived | Variant the name implies (free, a checkpoint date), copied to board_entry.variant where the source gave none. |
method | varchar | id | derived | How the mapping was made: exact, rule, manual, crosswalk. |
added | date | date | derived | Day the row was added to the curated file. |
provider_model_aliasA provider's native model id mapped to the canonical model: the crosswalk that resolves price and speed boards, where a row is a model on a provider.
9,264 rows. Key: file, line. Also as provider_model_alias.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
file | varchar | id | derived | File under data/curated/aliases/provider_model/. [key] |
line | integer | count | derived | Line of that file. [key] |
source_id | varchar | id | derived | Source that states the mapping. [joins source.id] |
provider | varchar | text | derived | Provider, as that source writes it. |
provider_model_id | varchar | text | derived | The provider's own id for the model. |
entity_id | varchar | id | derived | Canonical model id. [joins model.id] |
method | varchar | id | derived | How the mapping was made. |
added | date | date | derived | Day the row was added to the curated file. |
variant | varchar | text | derived | Column of data/curated/provider_model_alias that the schema does not declare yet; passed through as text. |
hf_repo | varchar | text | derived | Column of data/curated/provider_model_alias that the schema does not declare yet; passed through as text. |
status | varchar | text | derived | Column of data/curated/provider_model_alias that the schema does not declare yet; passed through as text. |
taskCanonical task categories (agent:memory_extraction, code-debugging): the entities of task boards such as OpenRouter's task classifications.
29 rows. Key: id. Also as task.csv.gz.
| column | type | unit | tier | description |
|---|---|---|---|---|
id | varchar | id | derived | Canonical task id. [key] |
display_name | varchar | text | derived | Name for people. |
group | varchar | id | derived | Task group (agent, code, data). |
source | varchar | id | derived | Where the row was seeded from. |
added | date | date | derived | Day the row was added to the curated file. |