Vending-Bench (minimum net worth)
Games · 2025-02-20
Minimum final net worth across repeated runs in the original Vending-Bench long-horizon vending business simulation.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3 Pro Preview | 3770 |
| Grok 4 | 3333 |
| GPT-5 | 2690 |
| GPT-5.1 | 1425 |
| Opus 4 | 1250 |
| Sonnet 4.5 | 1095 |
| Human (Vending-Bench) | 844 |
| Gemini 2.5 Pro Preview 03-25 | 692 |
| o3 | 569 |
| Claude 3.5 Sonnet (Oct 2024) | 476 |
| Gemini 1.5 Flash | 476 |
| Qwen3-235B-A22B | 453 |
| Sonnet 4 | 444 |
| Gemini 1.5 Pro | 439 |
| GPT-5 Mini | 422 |
| GPT-4o Mini | 421 |
| Kimi K2 Thinking | 419 |
| o3-mini | 369 |
| Gemini Robotics-ER 1.5 | 337 |
| gpt-oss-120b | 279 |
| Claude 3.7 Sonnet | 276 |
| Gemini 2.0 Pro Experimental | 274 |
| GPT-4o | 266 |
| Haiku 3.5 | 264 |
| Gemini 2.0 Flash | 157 |
| Llama 4 Maverick | -24.0 |