About BenchGecko
BenchGecko publishes datasets on the AI economy, including model prices, company figures, developer attention, GPU rental prices, and results from its own model tests. The files are refreshed daily, with Git history providing a time series for each file.
Key features
- 01Model prices and provider offers
- 02AI company figures with sources
- 03Daily developer attention data
- 04GPU rental price statistics
- 05Gecko Tests model results
- 06Daily snapshots in Git history
Use cases
- 01Compare model and provider offers
- 02Track reported AI company figures
- 03Analyze attention trends for models or topics
- 04Review GPU rental price trends
- 05Inspect model test answers and scores
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://benchgecko.ai/api/mcpclaude mcp add --transport http ai-benchgecko-benchgecko https://benchgecko.ai/api/mcp{"mcpServers":{"ai-benchgecko-benchgecko":{"type":"http","url":"https://benchgecko.ai/api/mcp"}}}Open in Cursor{"servers":{"ai-benchgecko-benchgecko":{"type":"http","url":"https://benchgecko.ai/api/mcp"}}}Open in VS Code{"mcpServers":{"ai-benchgecko-benchgecko":{"type":"http","url":"https://benchgecko.ai/api/mcp"}}}{"mcpServers":{"ai-benchgecko-benchgecko":{"serverUrl":"https://benchgecko.ai/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
9| Parameter | Type | Required | Description |
|---|---|---|---|
| limit | integer | No | |
| query | string | Yes | Model name, slug or lab, e.g. "claude opus", "gpt-5", "deepseek" |
Details
From the project READMEBenchGecko open datasets
Daily snapshots of the data BenchGecko collects and computes about the AI economy: model prices across providers, price changes, company figures with sources, developer attention (Mindshare), GPU rental prices and the results of BenchGecko's own model tests (Gecko Tests).
Files are refreshed every day by an automated job. Every day's version is in the git history, so the history of each file is a time series.
License
CC BY 4.0. Free to use, share and adapt, including commercially. Attribution required: "Source: BenchGecko" with a link to https://benchgecko.ai.
Cite as:
BenchGecko (2026). BenchGecko open datasets. https://github.com/BenchGecko/datasets · https://benchgecko.ai
Files
| File | What it is | Source |
|---|---|---|
prices/price-history.csv | Daily list price per model (USD per 1M tokens: input, output, cache read) since 2026-03-30 | OpenRouter catalog, recorded daily |
prices/provider-offers.csv | Every provider offer per model today: price, quantization, context, 1-day uptime | OpenRouter endpoints, daily |
prices/price-changes.csv | Dated events: new models, price changes that held (5%+), company updates | Computed by BenchGecko |
companies/company-facts.csv | Valuation, revenue, funding, headcount per AI company, each with as-of date and source URL when known | SEC filings, market close, press verified against the source page |
mindshare/mindshare-daily.csv | Daily attention per model, lab, agent, person and topic: Hacker News mentions, Wikipedia views, GitHub stars, 7-day share | Hacker News Algolia API, Wikimedia, GitHub |
gpu/gpu-rental-daily.csv | Median, 10th percentile and minimum on-demand rental price per GPU-hour for tracked AI chips | Vast.ai public marketplace |
gecko-tests/world-map.csv | World Map test: land or water answers on a 6° grid per model, scored against Natural Earth | BenchGecko test runs |
gecko-tests/<test>.json | Gecko Lab tests (who-are-you, tokenizer-tax, knowledge-horizon, same-model-different-host): questions, answer keys, every model answer and score | BenchGecko test runs |
Methodology for each test: https://benchgecko.ai/gecko-tests
Benchmark scores aggregated from third-party leaderboards are not redistributed here; see each leaderboard's own license.
Questions or corrections
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.