GAMUT: leaderboard
Factual completeness on 1,813 image-grounded questions, measured by the paper's headline importance-weighted Gamut score. Rubric checks reward complete answers and penalize contradictions; the score ranges from -100 to 100 percent.
Metric: Gamut Weighted Score (%). Source: arxiv.org. Saturation forecast: Around January 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 58.7 | #58 |
| 2 | Gemini 3 Flash | 57.9 | #99 |
| 3 | Claude Opus 4.8 | 53.1 | #37 |
| 4 | Gemini 2.5 Pro | 51.8 | #152 |
| 5 | Claude Opus 4.6 | 50.5 | #63 |
| 6 | GPT-5.4 | 42.6 | #79 |
| 7 | Claude Sonnet 4.6 | 41.5 | #90 |
| 8 | GPT-4o | 34.2 | #343 |
| 9 | Qwen 3 VL 235B A22B | 33 | #238 |
| 10 | Llama 4 Maverick | 18.3 | #464 |
| 11 | Llama 4 Scout | 14.1 | #656 |
| 12 | Llama 3.2 90B | 11.8 | #596 |
| 13 | Llama 3.2 11B | 5.1 | #1199 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=gamut · How It Works · Data refreshed daily, snapshot 2026-10-20.