Benchmarks · agentic
Vending-Bench Arena
Multi-agent Vending-Bench 2: models run competing vending machines at one location, may trade or collude, and are scored individually on final balance.
12 models · 12 results · data as of 2026-10-29 · andonlabs.com ↗
Leaderboard
GPT-6 Astra max
All results
12 rows| # | Model | Effort | Vending-Bench Arena |
|---|---|---|---|
| 1 | GPT-6 Astra | max | $12,400 |
| 2 | Opus 5.5 | max | $8,100 |
| 3 | GPT-5.5 | xhigh | $7,980 |
| 4 | GLM 5.3 | max | $7,800 |
| 5 | GPT-5.6 Sol | max | $7,400 |
| 6 | Opus 5 | max | $7,000 |
| 7 | Opus 4.7 | max | $5,838 |
| 8 | Sonnet 4.6 | max | $5,639 |
| 9 | Opus 4.8 | max | $5,188 |
| 10 | Fable 5 | max | $4,200 |
| 11 | Opus 4.6 | max | $4,053 |
| 12 | Kimi K3 | max | $3,200 |