Benchmarks · Agents & computer use · Coding
BiGym 2.0
Uses mean success on the nine core tasks. Excludes learned-policy baselines. Score: mean core-task success rate. Collected by Jake Boggs' benchmark dashboard.
9 models · 9 results · data as of 2026-11-15 · bigym2.github.io ↗
Leaderboard
GPT-6 Astra max
All results
9 rows| # | Model | Effort | BiGym 2.0 |
|---|---|---|---|
| 1 | GPT-6 Astra | max | 71.1% |
| 2 | Sonnet 5.5 | max | 61.0% |
| 3 | Opus 5.5 | high | 60.6% |
| 4 | GPT-6.1 Sol | max | 58.8% |
| 5 | GPT-6.1 Terra | max | 49.1% |
| 6 | GPT-6 Sol | max | 28.1% |
| 7 | Kimi K3.1 | max | 21.3% |
| 8 | DeepSeek V4.1 Flash 1017 | max | 13.1% |
| 9 | Haiku 5.5 | max | 8.3% |