AcuityBench (Conversational): leaderboard
Metric: Exact acuity match (%) in the conversational format on the 527 clear consensus cases: the model answers naturally without being asked for a label, a GPT-4.1 judge infers the communicated acuity level, mode of five samples at temperature 1; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Flash | 78 |
| 2 | GPT-5.4 | 77.2 |
| 3 | DeepSeek V3.1 | 76.8 |
| 4 | Gemini 2.5 Pro | 76.7 |
| 5 | Claude Sonnet 4.6 | 74.4 |
| 6 | Claude Opus 4.7 | 71.9 |
| 7 | GPT-4.1 | 71.4 |
| 8 | Claude Haiku 4.5 (20251001) | 70 |
| 9 | GPT-5 Mini | 67.7 |
| 10 | Qwen 2.5 72B Instruct Turbo | 66.8 |
| 11 | Llama 3.3 70B Instruct | 61.9 |
Interactive version: theaggregate.ai/benchmark?slug=acuitybench-conversational · How It Works · Data refreshed daily, snapshot 2026-10-07.