Vending-Bench (units sold)
Games · 2025-02-20
Mean units sold in the original Vending-Bench long-horizon vending business simulation.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok 4 | 4569 |
| Gemini 3 Pro Preview | 4200 |
| Sonnet 4.5 | 3099 |
| GPT-5 | 2471 |
| GPT-5.1 | 2105 |
| Claude 3.5 Sonnet (Oct 2024) | 1560 |
| Opus 4 | 1412 |
| o3 | 1363 |
| Kimi K2 Thinking | 1106 |
| Claude 3.7 Sonnet | 1050 |
| o3-mini | 831 |
| gpt-oss-120b | 698 |
| Sonnet 4 | 538 |
| Gemini Robotics-ER 1.5 | 537 |
| GPT-4o Mini | 473 |
| Qwen3-235B-A22B | 385 |
| Gemini 1.5 Pro | 375 |
| Gemini 2.5 Pro Preview 03-25 | 356 |
| Human (Vending-Bench) | 344 |
| GPT-5 Mini | 277 |
| GPT-4o | 258 |
| Gemini 2.0 Pro Experimental | 118 |
| Gemini 2.0 Flash | 104 |
| Gemini 1.5 Flash | 89.0 |
| Haiku 3.5 | 23.0 |
| Llama 4 Maverick | 0.0 |