CoffeeBench: leaderboard
Mean net income in dollars for the focal coffee roaster across three completed 90-day-or-bankruptcy seeds. Native run precision and exact checkpoint/provider-specific reasoning settings are preserved. The source off knob maps to different wire settings by provider; no blanket disabled-reasoning assumption is made.
Metric: Mean net income ($). Source: pub.sakana.ai. Saturation forecast: Around June 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 (Non-reasoning) | 2781.93 | #49 (Claude Opus 4.7) |
| 2 | Claude Sonnet 4.6 (Non-reasoning) | 2235.94 | #90 (Claude Sonnet 4.6) |
| 3 | Gemini 3.1 Pro (Preview) (Low) | 1695.05 | #58 (Gemini 3.1 Pro (Preview)) |
| 4 | GLM-5.1 (Non-reasoning) | 1597.12 | #97 (GLM-5.1) |
| 5 | Kimi K2.6 (Non-reasoning) | 453.93 | #104 (Kimi K2.6) |
| 6 | Claude Haiku 4.5 (Non-reasoning) | -630.04 | #280 (Claude Haiku 4.5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=coffeebench · How It Works · Data refreshed daily, snapshot 2026-10-20.