FactArena - Verdict Accuracy: leaderboard
FactArena (Lin et al., 2026) claim-verification accuracy (%): whether the model's final verdict on each complex HOVER or FEVEROUS claim, including arena-evolved harder claims, is correct. Higher is better.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 06-05) | 66.52 | #207 |
| 2 | O3 (2025-04-16) | 65.02 | #212 |
| 3 | GPT-4.1 | 64.32 | #492 |
| 4 | O4 Mini (2025-04-16) | 64.14 | #362 |
| 5 | Llama 4 Maverick Instruct | 60.91 | #1125 |
| 6 | GPT-4.5 (Preview) | 59.41 | #499 |
| 7 | DeepSeek V3 | 55.77 | #705 |
| 8 | Gemini 2.5 Flash | 55.51 | #452 |
| 9 | DeepSeek R1 | 55.14 | #467 |
| 10 | Grok 3 | 55.14 | #600 |
| 11 | Qwen 3 235B A22B | 55.13 | #622 |
| 12 | Grok 3 Mini Beta | 52.75 | #615 |
| 13 | Qwen 3 32B | 51.76 | #960 |
| 14 | GPT-4o | 50.95 | #741 |
| 15 | Claude Opus 4 (20250514) | 44.34 | #276 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=factarena-verdict-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-12.