Epoch AI - Vending Bench 2: leaderboard

Epoch AI mirror of Vending-Bench 2, a long-horizon agent benchmark for managing a simulated vending machine business.

Metric: Score. Source: epoch.ai. 69 models tracked.

Top models

#ModelScoreOverall rank
1gpt-6-astra_unknown15514.7
2gpt-6-sol_unknown14427.85
3gemini-4-argon_unknown13718.16
4claude-opus-5_unknown11181.87
5claude-sonnet-5-5_unknown11165.28
6claude-opus-4-7_unknown10936.76
7grok-4.7_unknown10536.83
8gpt-5.6-sol_unknown9619.37
9muse-wonder-1.0_unknown9432.18
10claude-opus-5-5_unknown9235.25

Interactive version: theaggregate.ai/benchmark?slug=epoch-ai-vending-bench-2 · How It Works · Data refreshed daily, snapshot 2026-10-20.