Benchmarks · agentic
Vending-Bench 2
Year-long simulated vending-machine business; agents handle suppliers, inventory, pricing, and refunds. Score is the final bank balance.
66 models · 71 results · data as of 2026-11-03 · andonlabs.com ↗
Leaderboard
Gemini 4 Carbon max
All results
71 rows| # | Model | Effort | Vending-Bench 2 |
|---|---|---|---|
| 1 | Gemini 4 Carbon | max | $16,484 |
| 2 | GPT-6 Astra | max | $15,515 |
| 3 | GPT-6 Sol | max | $14,428 |
| 4 | Gemini 4 Argon | max | $13,718 |
| 5 | Muse Wonder Full | max | $12,292 |
| 6 | Opus 5 | max | $11,182 |
| 7 | Sonnet 5.5 | max | $11,165 |
| 8 | Opus 4.7 | max | $10,937 |
| 9 | Grok 4.7 | max | $10,537 |
| 10 | GPT-5.6 Sol | max | $9,619 |
| 11 | Muse Wonder (Max) | max | $9,432 |
| 12 | Opus 5.5 | max | $9,235 |
| 13 | GPT-6.1 Terra | max | $9,107 |
| 14 | Grok 4.6 | max | $9,047 |
| 15 | GLM 5.2 | max | $8,314 |
| 16 | GLM 5.3 | max | $8,164 |
| 17 | Opus 4.6 | max | $8,018 |
| 18 | GPT-5.5 | xhigh | $7,524 |
| 19 | GPT-5.6 Terra | max | $7,343 |
| 20 | Sonnet 4.6 | max | $7,204 |
| 21 | Muse Spark 1.1 | max | $6,520 |
| 22 | Sonnet 5 | max | $6,378 |
| 23 | Kimi K2.6 | max | $6,205 |
| 24 | GPT-5.4 | xhigh | $6,144 |
| 25 | GPT-5.3 Codex | xhigh | $5,940 |
| 26 | Opus 4.8 | high | $5,787 |
| 27 | Fable 5 | high | $5,680 |
| 28 | GLM 5.1 | max | $5,634 |
| 29 | Gemini 3 Pro | high | $5,478 |
| 30 | Fable 5.1 | max | $5,422 |