MMT-Bench
The official leaderboard hasn’t scored a model released after DeepSeek-VL-7B-Chat (March 2024). Labs report their own numbers for 2 newer models, such as InternVL3.5-241B-A28B at 72.7%.
Vision-language models answer 31,325 multiple-choice questions across many multimodal tasks such as recognition, localization, reasoning and planning.
ICML 2024.
- Made by
- Shanghai AI Laboratory
- Released
- April 2024
- Kind
- General visual questions
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Last in a launch post
- October 2026
- Last model added
- October 2026
- Scores refreshed
- November 20th, 2026
In the authors’ words
Large Vision-Language Models (LVLMs) show significant strides in general-purpose multimodal applications such as visual dialogue and embodied navigation. However, existing multimodal evaluation benchmarks cover a limited number of multimodal tasks testing rudimentary capabilities, falling short in tracking LVLM development. In this study, we present MMT-Bench, a comprehensive benchmark designed to assess LVLMs across massive multimodal tasks requiring expert knowledge and deliberate visual recognition, localization, reasoning, and planning. MMT-Bench comprises 31,325 meticulously curated multi-choice visual questions from various multimodal scenarios such as vehicle driving and embodied navigation, covering 32 core meta-tasks and 162 subtasks in multimodal understanding. Due to its extensive task coverage, MMT-Bench enables the evaluation of LVLMs using a task map, facilitating the discovery of in- and out-of-domain tasks. Evaluation results involving 30 LVLMs such as the proprietary GPT-4V, GeminiProVision, and open-sourced InternVL-Chat, underscore the significant challenges posed by MMT-Bench. We anticipate that MMT-Bench will inspire the community to develop next-generation multimodal foundation models aimed at achieving general-purpose multimodal intelligence.
GPT4o (snapshot unspecified) tops the leaderboard with 65.5% Qwen-VL-Max leads with 62.4% among models released by March 2024
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Qwen-VL-Max, at 62.4%. This view shows only models we track, each at its best.
MiniCPM-V-4.7-35B-A3B leads with 75.5%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
The newest model with a score here came out in March 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 65.5 GPT4o (snapshot unspecified)
- 63.4 InternVL-Chat-v1.2-34B
- 62.4 Qwen-VL-Max
- 62.3
Qwen-VL-Plus
- 61.6 GeminiProVision
- 61.1 GPT-4V
- 60.8
LLaVA-v1.6-34B
- 55.7 XComposer2
- 54.8 BLIP2
- 54.2
Yi-VL-34B
- 53.4 Monkey-Chat
- 53.2
Yi-VL-6B
- 53.2
DeepSeek-VL-7B-Chat
- 53.0 llava-v1.6-vicuna-13b
- 52.7 TransCore-M
- 52.5 Qwen-VL-Chat
- 52.2 Claude3V_Haiku
- 52.1 XComposer
- 52.0
mPLUG-Owl2
- 51.8 RBDash-v1-13B
- 51.7
LLaVA-v1.5-13B
- 51.6 CogVLM-17B-Chat
- 51.5
ShareGPT4V-7B
- 51.1 LLaVA-NEXT-7B
- 51.1 LLaVA-v1.5-13B-XTuner
- 50.8 LLaVA-InternLM2-7B
- 50.2 LLaVA-v1.5-7B-XTuner
- 49.9 SharedCaptioner
- 49.7 LLaVA-InternLM-7B
- 49.5 llava-v1.5-7b
- 40.4 LLaMA-Adapter-v2-7B
- 38.6 VisualGLM-6B
- 31.7 Frequency Guess
- 1 GPT4o (snapshot unspecified) 65.5%
- 2 InternVL-Chat-v1.2-34B 63.4%
- 3 Qwen-VL-Max 62.4%
- 4
Qwen-VL-Plus 62.3%
- 5 GeminiProVision 61.6%
- 6 GPT-4V 61.1%
- 7
LLaVA-v1.6-34B 60.8%
- 8 XComposer2 55.7%
- 9 BLIP2 54.8%
- 10
Yi-VL-34B 54.2%
- 11 Monkey-Chat 53.4%
- 12
Yi-VL-6B 53.2%
- 13
DeepSeek-VL-7B-Chat 53.2%
- 14 llava-v1.6-vicuna-13b 53%
- 15 TransCore-M 52.7%
- 16 Qwen-VL-Chat 52.5%
- 17 Claude3V_Haiku 52.2%
- 18 XComposer 52.1%
- 19
mPLUG-Owl2 52%
- 20 RBDash-v1-13B 51.8%
- 21
LLaVA-v1.5-13B 51.7%
- 22 CogVLM-17B-Chat 51.6%
- 23
ShareGPT4V-7B 51.5%
- 24 LLaVA-NEXT-7B 51.1%
- 25 LLaVA-v1.5-13B-XTuner 51.1%
- 26 LLaVA-InternLM2-7B 50.8%
- 27 LLaVA-v1.5-7B-XTuner 50.2%
- 28 SharedCaptioner 49.9%
- 29 LLaVA-InternLM-7B 49.7%
- 30 llava-v1.5-7b 49.5%
- 31 LLaMA-Adapter-v2-7B 40.4%
- 32 VisualGLM-6B 38.6%
- 33 Frequency Guess 31.7%
Source: Official MMT-Bench leaderboard
- 1 Qwen-VL-Max 62.4%
- 2
Qwen-VL-Plus 62.3%
- 3 GPT-4V 61.1%
- 4
LLaVA-v1.6-34B 60.8%
- 5
Yi-VL-34B 54.2%
- 6 Monkey-Chat 53.4%
- 7
Yi-VL-6B 53.2%
- 8
DeepSeek-VL-7B-Chat 53.2%
- 9 llava-v1.6-vicuna-13b 53%
- 10 Qwen-VL-Chat 52.5%
- 11
mPLUG-Owl2 52%
- 12
LLaVA-v1.5-13B 51.7%
- 13 CogVLM-17B-Chat 51.6%
- 14
ShareGPT4V-7B 51.5%
- 15 llava-v1.5-7b 49.5%
Source: Official MMT-Bench leaderboard
- 75.5
MiniCPM-V-4.7-35B-A3B
- 72.7
InternVL3.5-241B-A28B
- 63.8
Nemotron Nano 12B v2 VL
- 62.4 Qwen-VL-Max
- 62.3
Qwen-VL-Plus
- 61.1 GPT-4V
- 60.8
LLaVA-v1.6-34B
- 54.2
Yi-VL-34B
- 53.4 Monkey-Chat
- 53.2
Yi-VL-6B
- 53.2
DeepSeek-VL-7B-Chat
- 53.0 llava-v1.6-vicuna-13b
- 52.5 Qwen-VL-Chat
- 52.0
mPLUG-Owl2
- 51.7
LLaVA-v1.5-13B
- 51.6 CogVLM-17B-Chat
- 51.5
ShareGPT4V-7B
- 49.5 llava-v1.5-7b
- 1
MiniCPM-V-4.7-35B-A3B 75.5%
- 2
InternVL3.5-241B-A28B 72.7%
- 3
Nemotron Nano 12B v2 VL 63.8%
- 4 Qwen-VL-Max 62.4%
- 5
Qwen-VL-Plus 62.3%
- 6 GPT-4V 61.1%
- 7
LLaVA-v1.6-34B 60.8%
- 8
Yi-VL-34B 54.2%
- 9 Monkey-Chat 53.4%
- 10
Yi-VL-6B 53.2%
- 11
DeepSeek-VL-7B-Chat 53.2%
- 12 llava-v1.6-vicuna-13b 53%
- 13 Qwen-VL-Chat 52.5%
- 14
mPLUG-Owl2 52%
- 15
LLaVA-v1.5-13B 51.7%
- 16 CogVLM-17B-Chat 51.6%
- 17
ShareGPT4V-7B 51.5%
- 18 llava-v1.5-7b 49.5%
Sources: Labs’ model cards and official MMT-Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT4o (snapshot unspecified) | 65.5% | Official leaderboard | ||
| 2 | InternVL-Chat-v1.2-34B | 63.4% | Official leaderboard | ||
| 3 | Qwen-VL-Max | 62.4% | Jan 2024 | Official leaderboard | |
| 4 | | 62.3% | Jan 2024 | Official leaderboard | |
| 5 | GeminiProVision | 61.6% | Official leaderboard | ||
| 6 | GPT-4V | 61.1% | Nov 2023 | Official leaderboard | |
| 7 | | 60.8% | Jan 2024 | Official leaderboard | |
| 8 | XComposer2 | 55.7% | Official leaderboard | ||
| 9 | BLIP2 | 54.8% | Official leaderboard | ||
| 10 | | 54.2% | Dec 2023 | Official leaderboard | |
| 11 | Monkey-Chat | 53.4% | Jan 2024 | Official leaderboard | |
| 12 | | 53.2% | Dec 2023 | Official leaderboard | |
| 13 | | 53.2% | Mar 2024 | Official leaderboard | |
| 14 | llava-v1.6-vicuna-13b | 53% | Jan 2024 | Official leaderboard | |
| 15 | TransCore-M | 52.7% | Official leaderboard | ||
| 16 | Qwen-VL-Chat | 52.5% | Aug 2023 | Official leaderboard | |
| 17 | Claude3V_Haiku | 52.2% | Official leaderboard | ||
| 18 | XComposer | 52.1% | Official leaderboard | ||
| 19 | | 52% | Nov 2023 | Official leaderboard | |
| 20 | RBDash-v1-13B | 51.8% | Official leaderboard | ||
| 21 | | 51.7% | Oct 2023 | Official leaderboard | |
| 22 | CogVLM-17B-Chat | 51.6% | Nov 2023 | Official leaderboard | |
| 23 | | 51.5% | Nov 2023 | Official leaderboard | |
| 24 | LLaVA-NEXT-7B | 51.1% | Official leaderboard | ||
| 25 | LLaVA-v1.5-13B-XTuner | 51.1% | Official leaderboard | ||
| 26 | LLaVA-InternLM2-7B | 50.8% | Official leaderboard | ||
| 27 | LLaVA-v1.5-7B-XTuner | 50.2% | Official leaderboard | ||
| 28 | SharedCaptioner | 49.9% | Official leaderboard | ||
| 29 | LLaVA-InternLM-7B | 49.7% | Official leaderboard | ||
| 30 | llava-v1.5-7b | 49.5% | Oct 2023 | Official leaderboard | |
| 31 | LLaMA-Adapter-v2-7B | 40.4% | Official leaderboard | ||
| 32 | VisualGLM-6B | 38.6% | Official leaderboard | ||
| 33 | Frequency Guess | 31.7% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Qwen-VL-Max | 62.4% | Jan 2024 | Official leaderboard | |
| 2 | | 62.3% | Jan 2024 | Official leaderboard | |
| 3 | GPT-4V | 61.1% | Nov 2023 | Official leaderboard | |
| 4 | | 60.8% | Jan 2024 | Official leaderboard | |
| 5 | | 54.2% | Dec 2023 | Official leaderboard | |
| 6 | Monkey-Chat | 53.4% | Jan 2024 | Official leaderboard | |
| 7 | | 53.2% | Dec 2023 | Official leaderboard | |
| 8 | | 53.2% | Mar 2024 | Official leaderboard | |
| 9 | llava-v1.6-vicuna-13b | 53% | Jan 2024 | Official leaderboard | |
| 10 | Qwen-VL-Chat | 52.5% | Aug 2023 | Official leaderboard | |
| 11 | | 52% | Nov 2023 | Official leaderboard | |
| 12 | | 51.7% | Oct 2023 | Official leaderboard | |
| 13 | CogVLM-17B-Chat | 51.6% | Nov 2023 | Official leaderboard | |
| 14 | | 51.5% | Nov 2023 | Official leaderboard | |
| 15 | llava-v1.5-7b | 49.5% | Oct 2023 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | 75.5% | Oct 2026 | Lab model card | ||
| 2 | 72.7% | Aug 2025 | Lab model card | ||
| 3 | 63.8% | Nov 2025 | Lab model card | ||
| 4 | Qwen-VL-Max | 62.4% | Jan 2024 | Official leaderboard | |
| 5 | 62.3% | Jan 2024 | Official leaderboard | ||
| 6 | GPT-4V | 61.1% | Nov 2023 | Official leaderboard | |
| 7 | 60.8% | Jan 2024 | Official leaderboard | ||
| 8 | 54.2% | Dec 2023 | Official leaderboard | ||
| 9 | Monkey-Chat | 53.4% | Jan 2024 | Official leaderboard | |
| 10 | 53.2% | Dec 2023 | Official leaderboard | ||
| 11 | 53.2% | Mar 2024 | Official leaderboard | ||
| 12 | llava-v1.6-vicuna-13b | 53% | Jan 2024 | Official leaderboard | |
| 13 | Qwen-VL-Chat | 52.5% | Aug 2023 | Official leaderboard | |
| 14 | 52% | Nov 2023 | Official leaderboard | ||
| 15 | 51.7% | Oct 2023 | Official leaderboard | ||
| 16 | CogVLM-17B-Chat | 51.6% | Nov 2023 | Official leaderboard | |
| 17 | 51.5% | Nov 2023 | Official leaderboard | ||
| 18 | llava-v1.5-7b | 49.5% | Oct 2023 | Official leaderboard |
Labs report scores for 3 models in their cards. Switch the scores above to rank them together.
Number one over time 3 models have held number one with claims included since its release in April 2024
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Sources: Official MMT-Bench leaderboard and labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official MMT-Bench leaderboard and labs’ model cards
One lab put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Oct 7, 2026 | 75.5 | MMT-Bench |
2 system cards report it, from 2 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 6, 2025 | | 63.8 | MMT-Bench (Val), reasoning on (Table 3). Reasoning off 65.5 | Capabilities |
| Aug 25, 2025 | | 72.7 | MMT (val) (Table 5) | Capabilities |
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy.
Qwen 3.6 Plus 86%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy.
Seed2.1 Pro 90.7%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy.
Qwen 3.8 Max 85.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Confidence check. Likely. Some facts rest on secondary sources. Every benchmark
Catalog record 1 warning 1 pass
- Pass Task access All tasks are public.
- Warning Confidence Likely. Some facts rest on secondary sources.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (33 entries), launch posts (1 score) and system cards (2 scores).
- Pass Models scored 18 models we track, plus 18 other entries on its leaderboard.
- Pass Newest score The newest is from October 2026, for MiniCPM-V-4.7-35B-A3B.
- Note Frontier models None of its models are on the capabilities index.
- Pass Headroom The best is 75.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 16 rows for models we track and 18 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.