chat benchmark
MM-MT-Bench Leaderboard
MM-MT-Bench is a multi-turn LLM-as-a-judge benchmark for evaluating multimodal instruction-tuned models.
Updated Sep 5, 2026
Models17
Model coverage17
MetricScore
EvidenceB
MM-MT-Bench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelMA | Score84.90 points | Percentile100.00% | Participants17 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score8.50 points | Percentile93.75% | Participants17 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score8.50 points | Percentile87.50% | Participants17 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score8.40 points | Percentile81.25% | Participants17 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score8.30 points | Percentile75.00% | Participants17 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score8.10 points | Percentile68.75% | Participants17 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score8.00 points | Percentile62.50% | Participants17 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score7.90 points | Percentile56.25% | Participants17 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score7.70 points | Percentile50.00% | Participants17 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score7.70 points | Percentile43.75% | Participants17 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score7.50 points | Percentile37.50% | Participants17 | EvidenceC | Evaluated |
| Rank12 | ModelMA | Score0.74 points | Percentile31.25% | Participants17 | EvidenceC | Evaluated |
| Rank13 | ModelMA | Score0.61 points | Percentile25.00% | Participants17 | EvidenceC | Evaluated |
| Rank14 | ModelMA | Score0.08 points | Percentile18.75% | Participants17 | EvidenceC | Evaluated |
| Rank15 | ModelMA | Score0.08 points | Percentile12.50% | Participants17 | EvidenceC | Evaluated |
| Rank16 | ModelMA | Score0.08 points | Percentile6.25% | Participants17 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score0.06 points | Percentile0.00% | Participants17 | EvidenceC | Evaluated |
MM-MT-Bench Highlights
The leading models and scores on this benchmark.
MM-MT-Bench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for MM-MT-Bench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is MM-MT-Bench?
What MM-MT-Bench measures and how its scores work.
MM-MT-Bench evaluates multimodal instruction-tuned models in multi-turn dialogues and zero-shot open-ended question answering.
It measures the ability to follow user instructions in multi-turn dialogues and answer open-ended questions, reported as Score in points.
Scores are shown in points. This benchmark is not independently verified and has an evidence level of B.
- Family
- MM-MT-Bench
- Modality
- multimodal
- Primary category
- chat
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MM-MT-Bench.
Which model scores highest on MM-MT-Bench?
Mistral Large 3 is currently ranked first with 84.90 points.
What are the top three models on MM-MT-Bench?
The current leaders are Mistral Large 3 (84.90 points), Qwen3 VL 235B A22B Thinking (8.50 points), and Qwen3 VL 235B A22B Instruct (8.50 points).
Which MM-MT-Bench model has the lowest official input price?
MiniStral 3 (14B Instruct 2512) has the lowest matched official input price at $0.10 input / $0.10 output per 1M tokens.
Which models are fastest among MM-MT-Bench results?
The fastest matched records are Qwen3 VL 4B Thinking (8.77 tok/s via DeepInfra), Pixtral Large (0.10 tok/s via Mistral AI), and Pixtral-12B (0.10 tok/s via Mistral AI).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.