NEVU - Macro-F1: leaderboard
Metric: Macro-F1 (%) over the directed Level-1 value labels present in the gold annotations on NEVU's sampled test subset, all four unit levels pooled, shared prompt and actor-centric JSON output, long units handled map-reduce; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 39.87 | #145 |
| 2 | Claude Opus 4.5 | 37.25 | #79 |
| 3 | Claude Sonnet 4.5 | 34.73 | #138 |
| 4 | GPT-5.2 | 32 | #105 |
| 5 | O3 | 31.22 | #121 |
| 6 | GPT-4.1 | 31.12 | #240 |
| 7 | Claude Haiku 4.5 | 30.11 | #271 |
| 8 | Ministral-3-8B-Instruct-2512 | 17.17 | #644 |
| 9 | Qwen 3 8B | 9.35 | #667 |
| 10 | Llama 3.1 8B Instruct | 5.94 | #1018 |
| 11 | Phi-3.5-mini-instruct | 5.73 | #1148 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=nevu-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.