LLM Stats (MuirBench): leaderboard
12 multi-image reasoning tasks requiring comparative analysis, temporal ordering, and relational reasoning across multiple related images.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B (Thinking) | 80.3 | #294 (Qwen 3 VL 32B) |
| 2 | Qwen 3 VL 235B A22B (Thinking) | 80.1 | #238 (Qwen 3 VL 235B A22B) |
| 3 | Seed 1.8 | 78.7 | #139 |
| 4 | Seed 2.0 Mini | 78 | #276 |
| 5 | Qwen 3 VL 30B A3B (Thinking) | 77.6 | #351 (Qwen 3 VL 30B A3B) |
| 6 | Qwen 3 VL 8B (Thinking) | 76.8 | |
| 7 | Qwen 3 VL 4B (Thinking) | 75 | #480 (Qwen 3 VL 4B) |
| 8 | Qwen 3 VL 32B Instruct | 72.8 | #283 |
| 9 | Qwen 3 VL 235B A22B Instruct | 72.8 | #273 |
| 10 | Qwen 3 VL 8B Instruct | 64.4 | #420 |
| 11 | Qwen 3 VL 4B Instruct | 63.8 | #515 |
| 12 | Qwen 3 VL 30B A3B Instruct | 62.9 | #384 |
| 13 | Qwen2.5-Omni-7B | 59.2 | #624 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-muirbench · How It Works · Data refreshed daily, snapshot 2026-10-20.