multimodal benchmark
MuirBench
A comprehensive benchmark for robust multi-image understanding capabilities of multimodal LLMs. Consists of 12 diverse multi-image tasks involving 10 categories of multi-image relations (e.g., multiview, temporal relations, narrative, complementary). Comprises 11,264 images and 2,600 multiple-choice questions created in a pairwise manner, where each standard instance is paired with an unanswerable variant for reliable assessment.
Updated Aug 7, 2026
Published models11
Registry coverage11
MetricScore
EvidenceB
MuirBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 80.3% | 100.0% | 11 | C | |
| 2 | AC | 80.1% | 90.0% | 11 | C | |
| 3 | AC | 77.6% | 80.0% | 11 | C | |
| 4 | AC | 76.8% | 70.0% | 11 | C | |
| 5 | AC | 75.0% | 60.0% | 11 | C | |
| 6 | AC | 72.8% | 50.0% | 11 | C | |
| 7 | AC | 72.8% | 40.0% | 11 | C | |
| 8 | AC | 64.4% | 30.0% | 11 | C | |
| 9 | AC | 63.8% | 20.0% | 11 | C | |
| 10 | AC | 62.9% | 10.0% | 11 | C | |
| 11 | AC | 59.2% | 0.0% | 11 | C |
Score distribution
Top published rows on the benchmark's original scale.
MuirBench highlights
The top published results on this benchmark's own scale.
What is MuirBench?
Definition and scoring fields from the benchmark registry.
A comprehensive benchmark for robust multi-image understanding capabilities of multimodal LLMs. Consists of 12 diverse multi-image tasks involving 10 categories of multi-image relations (e.g., multiview, temporal relations, narrative, complementary). Comprises 11,264 images and 2,600 multiple-choice questions created in a pairwise manner, where each standard instance is paired with an unanswerable variant for reliable assessment.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- MuirBench
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- muirbench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about MuirBench.
Which model scores highest on MuirBench?
Qwen3 VL 32B Thinking is currently ranked first with 80.3%.
What does MuirBench measure?
A comprehensive benchmark for robust multi-image understanding capabilities of multimodal LLMs. Consists of 12 diverse multi-image tasks involving 10 categories of multi-image relations (e.g., multiview, temporal relations, narrative, complementary). Comprises 11,264 images and 2,600 multiple-choice questions created in a pairwise manner, where each standard instance is paired with an unanswerable variant for reliable assessment.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
11 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.