Rankings · rolling 7 days

What agents actually run.

Request counts per model, measured from BlockRun's gateway request log over the last 7 days — not survey data, not self-reported. Every paid call behind these numbers settled in USDC on-chain.

#modelrequeststokensshare

1GPT-5.4 Miniopenai5.6M8.9B90.7%2GPT-4oopenai458.1K1.5B7.5%3GPT-5.5openai35.2K654.2M0.6%4GPT-4o Miniopenai14K7.2M0.2%5Claude Fable 5anthropic12.5K211.8M0.2%6GPT-4.1 Miniopenai7.5K5.1M0.1%7StepFun Step 3.7 Flash (Free)nvidiaFree6.5K112.3M0.1%8Claude Opus 4.8anthropic5.4K32.9M0.1%9GPT-5.6 Terraopenai4.7K348.1M0.1%10Nemotron 3 Nano Omni (Free)nvidiaFree3.9K53.7M0.1%11GPT-5 Miniopenai3.5K2.9M0.1%12Mistral Nemotron (Free)nvidiaFree3.4K19.8M0.1%13Nemotron Nano 9B v2 (Free)nvidiaFree2.5K15M0.0%14Claude Opus 4.7anthropic2.4K10.4M0.0%15Claude Opus 5anthropic2.3K49.2M0.0%16GPT-5.4openai2.1K20.5M0.0%17GPT-5.6 Solopenai1.1K122.7M0.0%18Claude Sonnet 5anthropic1.1K18M0.0%19GPT-5.6 Lunaopenai1K45.1M0.0%20Gemini 2.5 Flashgoogle7215.7M0.0%

Chat requests served by the gateway, rolling 7 days, refreshed hourly. Aliases of the same model are merged; retired models are excluded. Token column counts input + output.

Browse all models Watch the live feed