PhenoBench-LLM - All-Task Win Rate: leaderboard
PhenoBench-LLM (arXiv 2609.06080v1; release file win_rates.csv): pairwise win rate of 14 language models over all available head-to-heads on the validation tasks each model ran (22 to 40 of 40), so coverage differs by model. Native task metrics are compared only within a task; percent, higher is better.
Metric: Pairwise win rate (%; all available head-to-heads, 22-40 tasks per model). Source: github.com. Saturation forecast: Around February 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.7 Flash | 79 | #24 |
| 2 | Claude Sonnet 5 | 69.5 | #65 |
| 3 | Claude Opus 4.8 | 68.18 | #39 |
| 4 | Gemini 3.1 Pro (Preview) | 65.34 | #64 |
| 5 | GPT-5.6 Luna | 55.41 | #79 |
| 6 | Gemini 3.1 Flash Lite | 54.98 | #200 |
| 7 | Grok 4.3 | 53.03 | #204 |
| 8 | Nova Pro | 46.54 | #594 |
| 9 | Claude Haiku 4.5 | 44.48 | #287 |
| 10 | Llama 4 Maverick | 43.4 | #486 |
| 11 | DeepSeek V4 Pro | 36.07 | #112 |
| 12 | Llama 3.3 70B | 31.82 | #578 |
| 13 | GPT-5.4 Nano | 29.76 | #359 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=phenobench-llm-all-task-win-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.