CAME-Bench - Medium: leaderboard
Metric: Answer-set macro F1 (%; 168 questions on six trajectories averaging 137K tokens; full context, left-truncated to the model's window, gpt-4.1-mini judge). Source: arxiv.org. Saturation forecast: Around December 2026. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 Mini (Medium) | 56.6 |
| 2 | GPT-4.1 Mini | 36.2 |
| 3 | Qwen 3 235B A22B 2507 Instruct | 15.2 |
| 4 | GPT-4o Mini | 7.5 |
| 5 | DeepSeek V3.1 | 1 |
Interactive version: theaggregate.ai/benchmark?slug=came-bench-medium · How It Works · Data refreshed daily, snapshot 2026-09-26.