Vending-Bench 2
Games · 2025-11-18
Final bank-balance score in a simulated vending business benchmark.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 11182 |
| Opus 4.7 | 10937 |
| GPT-5.6 Sol | 9619 |
| GLM-5.2 | 8314 |
| Opus 4.6 | 8018 |
| GPT-5.5 | 7524 |
| GPT-5.6 Terra | 7343 |
| Sonnet 4.6 | 7204 |
| Sonnet 5 | 6378 |
| Kimi K2.6 | 6205 |
| GPT-5.4 | 6144 |
| GPT-5.3-Codex | 5940 |
| Opus 4.8 | 5787 |
| Claude Fable 5 | 5680 |
| GLM-5.1 | 5634 |
| Gemini 3 Pro Preview | 5478 |
| Gemini 3.5 Flash | 5396 |
| Qwen3.6 Plus (2026-04-02) | 5115 |
| Kimi K2.7 Code | 5083 |
| Opus 4.5 | 4967 |
| Grok 4.20 | 4663 |
| GLM-5 | 4432 |
| Qwen3.6-Max-Preview | 4254 |
| GPT-5.6 Luna | 4095 |
| GPT-5.2 | 3952 |
| Sonnet 4.5 | 3839 |
| Gemini 3 Flash Preview | 3635 |
| DeepSeek-V4-Pro | 3285 |
| MiniMax M3 | 2158 |
| GPT-5.1 | 1473 |
| Grok 4.1 Fast | 1107 |
| DeepSeek-V3.2-Exp | 1034 |
| DeepSeek-V3.2 | 1034 |
| Gemini 2.5 Flash | 549 |
| Qwen3.5 35B A3B | 463 |
| Qwen3.5-Flash | 463 |
| Qwen3 Max (rolling alias) | 71.6 |
| Qwen3-Max (2025-09-23) | 71.6 |
| Qwen3 235B A22B Thinking 2507 | -11.3 |
| MiniMax M2.5 | -23.2 |