Multi-Crit
Multimodal judges compare responses under several criteria. Human judgments provide the reference for each criterion and reveal trade-offs between criteria.
The November 2025 paper compares GPT, Gemini, Claude and open models. OpenAI o4-mini led the verifiable reasoning table at 53.17% pluralistic accuracy.
- Made by
- Multi-Crit authors
- Released
- November 2025
- Kind
- General visual questions
- Level
- University
- 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
- Strongest model tested
- GPT-5 #71 on the capability index
- Last model added
- October 2025
- Scores refreshed
- November 20th, 2026
o4-mini leads with 43% among models released by October 2025 o4-mini leads with 43% among models released by October 2025
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. This is the plain mean of pluralistic accuracy on open-ended and verifiable reasoning prompts. 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. This view shows only models we track, each at its best.
The newest model with a score here came out in October 2025, 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
- 43.0 o4-mini
- 37.8 o3
- 37.5 GPT-5
- 36.4 GPT-4o (snapshot unspecified)
- 35.0 Gemini-2.5-Pro (snapshot unspecified)
- 34.1 Claude 3.7 Sonnet
- 33.9
InternVL3.5-38B
- 33.2 MiMo-VL-7B
- 31.4
Gemini 2.5 Flash
- 30.5
Qwen2.5-VL-72B
- 29.2
InternVL3-78B
- 28.8
InternVL3.5-8B
- 26.5
InternVL3-8B
- 23.6 GLM-4.1V-9B-Thinking
- 22.1
MiniCPM-V 4.5
- 19.7
Eagle2.5-8B
- 18.4 R1-Reward-7B
- 17.5 LLaVA-Critic-R1-7B
- 17.5
Qwen3-VL-8B-Instruct
- 15.0 UnifiedReward-7B
- 14.1 Molmo-7B
- 13.0
Qwen2.5-VL-7B
- 12.6 LLaVA-Critic-7B
- 11.5
Llama 3.2 11B
- 10.5
LLaVA-OneVision 7B
- 1 o4-mini 43%
- 2 o3 37.8%
- 3 GPT-5 37.5%
- 4 GPT-4o (snapshot unspecified) 36.4%
- 5 Gemini-2.5-Pro (snapshot unspecified) 35%
- 6 Claude 3.7 Sonnet 34.1%
- 7
InternVL3.5-38B 33.9%
- 8 MiMo-VL-7B 33.2%
- 9
Gemini 2.5 Flash Sep 2025 31.4%
- 10
Qwen2.5-VL-72B 30.5%
- 11
InternVL3-78B 29.2%
- 12
InternVL3.5-8B 28.8%
- 13
InternVL3-8B 26.5%
- 14 GLM-4.1V-9B-Thinking 23.6%
- 15
MiniCPM-V 4.5 22.1%
- 16
Eagle2.5-8B 19.7%
- 17 R1-Reward-7B 18.4%
- 18 LLaVA-Critic-R1-7B 17.5%
- 19
Qwen3-VL-8B-Instruct 17.5%
- 20 UnifiedReward-7B 15%
- 21 Molmo-7B 14.1%
- 22
Qwen2.5-VL-7B 13%
- 23 LLaVA-Critic-7B 12.6%
- 24
Llama 3.2 11B 11.5%
- 25
LLaVA-OneVision 7B 10.5%
Source: The Multi-Crit paper
- 43.0 o4-mini
- 37.8 o3
- 37.5 GPT-5
- 34.1 Claude 3.7 Sonnet
- 33.9
InternVL3.5-38B
- 31.4
Gemini 2.5 Flash
- 30.5
Qwen2.5-VL-72B
- 29.2
InternVL3-78B
- 28.8
InternVL3.5-8B
- 26.5
InternVL3-8B
- 23.6 GLM-4.1V-9B-Thinking
- 22.1
MiniCPM-V 4.5
- 19.7
Eagle2.5-8B
- 17.5
Qwen3-VL-8B-Instruct
- 13.0
Qwen2.5-VL-7B
- 11.5
Llama 3.2 11B
- 10.5
LLaVA-OneVision 7B
- 1 o4-mini 43%
- 2 o3 37.8%
- 3 GPT-5 37.5%
- 4 Claude 3.7 Sonnet 34.1%
- 5
InternVL3.5-38B 33.9%
- 6
Gemini 2.5 Flash Sep 2025 31.4%
- 7
Qwen2.5-VL-72B 30.5%
- 8
InternVL3-78B 29.2%
- 9
InternVL3.5-8B 28.8%
- 10
InternVL3-8B 26.5%
- 11 GLM-4.1V-9B-Thinking 23.6%
- 12
MiniCPM-V 4.5 22.1%
- 13
Eagle2.5-8B 19.7%
- 14
Qwen3-VL-8B-Instruct 17.5%
- 15
Qwen2.5-VL-7B 13%
- 16
Llama 3.2 11B 11.5%
- 17
LLaVA-OneVision 7B 10.5%
Source: The Multi-Crit paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o4-mini | 43% | Apr 2025 | Official leaderboard | |
| 2 | o3 | 37.8% | Apr 2025 | Official leaderboard | |
| 3 | GPT-5 | 37.5% | Aug 2025 | Official leaderboard | |
| 4 | GPT-4o (snapshot unspecified) | 36.4% | Official leaderboard | ||
| 5 | Gemini-2.5-Pro (snapshot unspecified) | 35% | Official leaderboard | ||
| 6 | Claude 3.7 Sonnet | 34.1% | Feb 2025 | Official leaderboard | |
| 7 | | 33.9% | Aug 2025 | Official leaderboard | |
| 8 | MiMo-VL-7B | 33.2% | Official leaderboard | ||
| 9 | | 31.4% | Sep 2025 | Official leaderboard | |
| 10 | | 30.5% | Jan 2025 | Official leaderboard | |
| 11 | | 29.2% | Apr 2025 | Official leaderboard | |
| 12 | | 28.8% | Aug 2025 | Official leaderboard | |
| 13 | | 26.5% | Apr 2025 | Official leaderboard | |
| 14 | GLM-4.1V-9B-Thinking | 23.6% | Jul 2025 | Official leaderboard | |
| 15 | | 22.1% | Aug 2025 | Official leaderboard | |
| 16 | | 19.7% | Apr 2025 | Official leaderboard | |
| 17 | R1-Reward-7B | 18.4% | Official leaderboard | ||
| 18 | LLaVA-Critic-R1-7B | 17.5% | Official leaderboard | ||
| 19 | | 17.5% | Oct 2025 | Official leaderboard | |
| 20 | UnifiedReward-7B | 15% | Official leaderboard | ||
| 21 | Molmo-7B | 14.1% | Official leaderboard | ||
| 22 | | 13% | Jan 2025 | Official leaderboard | |
| 23 | LLaVA-Critic-7B | 12.6% | Official leaderboard | ||
| 24 | | 11.5% | Sep 2024 | Official leaderboard | |
| 25 | | 10.5% | Aug 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o4-mini | 43% | Apr 2025 | Official leaderboard | |
| 2 | o3 | 37.8% | Apr 2025 | Official leaderboard | |
| 3 | GPT-5 | 37.5% | Aug 2025 | Official leaderboard | |
| 4 | Claude 3.7 Sonnet | 34.1% | Feb 2025 | Official leaderboard | |
| 5 | | 33.9% | Aug 2025 | Official leaderboard | |
| 6 | | 31.4% | Sep 2025 | Official leaderboard | |
| 7 | | 30.5% | Jan 2025 | Official leaderboard | |
| 8 | | 29.2% | Apr 2025 | Official leaderboard | |
| 9 | | 28.8% | Aug 2025 | Official leaderboard | |
| 10 | | 26.5% | Apr 2025 | Official leaderboard | |
| 11 | GLM-4.1V-9B-Thinking | 23.6% | Jul 2025 | Official leaderboard | |
| 12 | | 22.1% | Aug 2025 | Official leaderboard | |
| 13 | | 19.7% | Apr 2025 | Official leaderboard | |
| 14 | | 17.5% | Oct 2025 | Official leaderboard | |
| 15 | | 13% | Jan 2025 | Official leaderboard | |
| 16 | | 11.5% | Sep 2024 | Official leaderboard | |
| 17 | | 10.5% | Aug 2024 | Official leaderboard |
Rankings by part
We read 18 parts from the benchmark’s paper in 2 groups: open-ended and verifiable reasoning. Each ranks models on one part of the test. o4-mini leads 11 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Pluralistic
17 models we track have a score here.
- 1 o4-mini 32.8%
- 2 Claude 3.7 Sonnet ↑2 31.8%
- 3 o3 ↓1 31.1%
- 4
InternVL3.5-38B ↑1 30.4%
- 5 GPT-5 ↓2 29.8%
- 6
InternVL3-78B ↑2 29.1%
- 7
Qwen2.5-VL-72B 28.4%
- 8
InternVL3-8B ↑2 26.1%
- 9
Gemini 2.5 Flash Sep 2025 ↓3 25.4%
- 10
InternVL3.5-8B ↓1 25.1%
Conflict
17 models we track have a score here.
- 1 o4-mini 43.1%
- 2 o3 42.7%
- 3 Claude 3.7 Sonnet ↑1 42.3%
- 4 GPT-5 ↓1 38.5%
- 5
Qwen2.5-VL-72B ↑2 35.5%
- 6
InternVL3.5-38B ↓1 33.7%
- 7
InternVL3-78B ↑1 32.5%
- 8
InternVL3.5-8B ↑1 32.3%
- 9
Gemini 2.5 Flash Sep 2025 ↓3 31.3%
- 10 GLM-4.1V-9B-Thinking ↑1 30.5%
Tradeoff
17 models we track have a score here.
- 1 o4-mini 64.6%
- 2 Claude 3.7 Sonnet ↑2 64.1%
- 3
InternVL3.5-38B ↑2 64.1%
- 4 o3 ↓2 62.6%
- 5 GPT-5 ↓2 62.6%
- 6
InternVL3.5-8B ↑3 62.1%
- 7
Gemini 2.5 Flash Sep 2025 ↓1 62.1%
- 8
InternVL3-8B ↑2 61.2%
- 9
Qwen2.5-VL-72B ↓2 60.7%
- 10
InternVL3-78B ↓2 56.3%
Completeness
17 models we track have a score here.
- 1 o4-mini 76.1%
- 2
InternVL3-78B ↑6 73%
- 3 Claude 3.7 Sonnet ↑1 71.7%
- 4
InternVL3.5-38B ↑1 71.7%
- 5 GPT-5 ↓2 69.9%
- 6
Qwen2.5-VL-72B ↑1 69.5%
- 7 o3 ↓5 66.4%
- 8
InternVL3-8B ↑2 64.6%
- 9
Gemini 2.5 Flash Sep 2025 ↓3 64.6%
- 10
Eagle2.5-8B ↑3 64.2%
Grounding
17 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑3 74.5%
- 2 o3 74.5%
- 3 o4-mini ↓2 74.5%
- 4 GPT-5 ↓1 69.6%
- 5
InternVL3-78B ↑3 65.6%
- 6
InternVL3.5-38B ↓1 65.2%
- 7
Qwen2.5-VL-72B 63.4%
- 8
InternVL3-8B ↑2 63%
- 9
Gemini 2.5 Flash Sep 2025 ↓3 60.7%
- 10
InternVL3.5-8B ↓1 59.4%
Hallucination
17 models we track have a score here.
- 1 GPT-5 ↑2 75.7%
- 2 o3 72.9%
- 3
Gemini 2.5 Flash Sep 2025 ↑3 66.3%
- 4 o4-mini ↓3 65.8%
- 5 Claude 3.7 Sonnet ↓1 63.5%
- 6
InternVL3.5-38B ↓1 63%
- 7 GLM-4.1V-9B-Thinking ↑4 61.3%
- 8
Qwen3-VL-8B-Instruct ↑6 59.1%
- 9
Qwen2.5-VL-72B ↓2 58.6%
- 10
InternVL3-78B ↓2 56.4%
Expressiveness
17 models we track have a score here.
- 1
Qwen2.5-VL-72B ↑6 71%
- 2 o4-mini ↓1 69.4%
- 3
InternVL3.5-8B ↑6 69.4%
- 4
InternVL3-78B ↑4 68.9%
- 5
InternVL3-8B ↑5 67.9%
- 6 GLM-4.1V-9B-Thinking ↑5 67.4%
- 7 Claude 3.7 Sonnet ↓3 66.8%
- 8
MiniCPM-V 4.5 ↑4 65.8%
- 9
Qwen2.5-VL-7B ↑6 64.1%
- 10
InternVL3.5-38B ↓5 63.7%
Clarity
17 models we track have a score here.
- 1 o3 ↑1 68.8%
- 2 GPT-5 ↑1 68.8%
- 3 o4-mini ↓2 62.5%
- 4
InternVL3.5-38B ↑1 61.9%
- 5 Claude 3.7 Sonnet ↓1 60.2%
- 6
InternVL3-78B ↑2 59.7%
- 7
InternVL3.5-8B ↑2 58.5%
- 8
Gemini 2.5 Flash Sep 2025 ↓2 57.4%
- 9
Qwen2.5-VL-72B ↓2 56.8%
- 10
InternVL3-8B 56.3%
Avg
17 models we track have a score here.
- 1 o4-mini 69.7%
- 2 o3 69.2%
- 3 GPT-5 68.5%
- 4 Claude 3.7 Sonnet 67.4%
- 5
InternVL3.5-38B 65.1%
- 6
InternVL3-78B ↑2 64.7%
- 7
Qwen2.5-VL-72B 63.8%
- 8
Gemini 2.5 Flash Sep 2025 ↓2 62.5%
- 9
InternVL3-8B ↑1 61.6%
- 10
InternVL3.5-8B ↓1 61%
Pluralistic
17 models we track have a score here.
- 1 o4-mini 53.2%
- 2 GPT-5 ↑1 45.2%
- 3 o3 ↓1 44.4%
- 4
InternVL3.5-38B ↑1 37.3%
- 5
Gemini 2.5 Flash Sep 2025 ↑1 37.3%
- 6 Claude 3.7 Sonnet ↓2 36.5%
- 7
Qwen2.5-VL-72B 32.5%
- 8
InternVL3.5-8B ↑1 32.5%
- 9
InternVL3-78B ↓1 29.4%
- 10
InternVL3-8B 27%
Conflict
17 models we track have a score here.
- 1 o4-mini 65.8%
- 2 o3 62.3%
- 3 GPT-5 56.6%
- 4
Gemini 2.5 Flash Sep 2025 ↑2 53%
- 5 Claude 3.7 Sonnet ↓1 52%
- 6
InternVL3.5-38B ↓1 47.7%
- 7
Qwen2.5-VL-72B 45.9%
- 8
InternVL3-78B 39.5%
- 9
InternVL3-8B ↑1 39.5%
- 10
InternVL3.5-8B ↓1 39.1%
Tradeoff
17 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑3 83.5%
- 2 o4-mini ↓1 83.5%
- 3 o3 ↓1 82.6%
- 4 GPT-5 ↓1 78.9%
- 5
Qwen2.5-VL-72B ↑2 77.1%
- 6
MiniCPM-V 4.5 ↑6 75.2%
- 7
InternVL3.5-38B ↓2 75.2%
- 8
Gemini 2.5 Flash Sep 2025 ↓2 73.4%
- 9
InternVL3.5-8B 69.7%
- 10
InternVL3-78B ↓2 67.9%
Grounding
17 models we track have a score here.
- 1 o4-mini 77.8%
- 2 Claude 3.7 Sonnet ↑2 74.1%
- 3 GPT-5 74.1%
- 4
InternVL3.5-38B ↑1 71.6%
- 5 o3 ↓3 67.9%
- 6
Gemini 2.5 Flash Sep 2025 67.5%
- 7
InternVL3.5-8B ↑2 65.4%
- 8
InternVL3-8B ↑2 64.2%
- 9
InternVL3-78B ↓1 63%
- 10
Qwen2.5-VL-72B ↓3 60.5%
Logic
17 models we track have a score here.
- 1 o4-mini 75.6%
- 2 o3 70.9%
- 3 GPT-5 70.9%
- 4
InternVL3.5-38B ↑1 70.9%
- 5
InternVL3-78B ↑3 69.8%
- 6
MiniCPM-V 4.5 ↑6 66.3%
- 7
InternVL3-8B ↑3 65.1%
- 8
InternVL3.5-8B ↑1 64%
- 9
Gemini 2.5 Flash Sep 2025 ↓3 64%
- 10 Claude 3.7 Sonnet ↓6 62.8%
Hallucination
17 models we track have a score here.
- 1 o3 ↑1 84.2%
- 2 GPT-5 ↑1 82.8%
- 3 o4-mini ↓2 79.3%
- 4 Claude 3.7 Sonnet 65.5%
- 5 GLM-4.1V-9B-Thinking ↑6 65.5%
- 6
Qwen3-VL-8B-Instruct ↑8 65.5%
- 7
Gemini 2.5 Flash Sep 2025 ↓1 64.9%
- 8
InternVL3-8B ↑2 58.6%
- 9
Qwen2.5-VL-72B ↓2 56.9%
- 10
MiniCPM-V 4.5 ↑2 56.9%
Exploration
17 models we track have a score here.
- 1 o3 ↑1 84.6%
- 2 o4-mini ↓1 83.5%
- 3 GPT-5 81.3%
- 4
InternVL3.5-38B ↑1 78%
- 5
Gemini 2.5 Flash Sep 2025 ↑1 78%
- 6
InternVL3.5-8B ↑3 71.4%
- 7
Qwen2.5-VL-72B 68.1%
- 8 Claude 3.7 Sonnet ↓4 68.1%
- 9
InternVL3-8B ↑1 67%
- 10
MiniCPM-V 4.5 ↑2 65.9%
Efficiency
17 models we track have a score here.
- 1 o4-mini 88.1%
- 2 o3 81.7%
- 3 GPT-5 78%
- 4
Gemini 2.5 Flash Sep 2025 ↑2 78%
- 5 Claude 3.7 Sonnet ↓1 76.2%
- 6
InternVL3-8B ↑4 76.2%
- 7
Qwen2.5-VL-72B 75.2%
- 8
InternVL3-78B 74.3%
- 9
InternVL3.5-38B ↓4 73.4%
- 10
InternVL3.5-8B ↓1 71.6%
Avg
17 models we track have a score here.
- 1 o4-mini 80.8%
- 2 o3 77.9%
- 3 GPT-5 77.4%
- 4
Gemini 2.5 Flash Sep 2025 ↑2 70.5%
- 5
InternVL3.5-38B 69.8%
- 6 Claude 3.7 Sonnet ↓2 69.3%
- 7
InternVL3-8B ↑3 66.2%
- 8
InternVL3.5-8B ↑1 65.8%
- 9
InternVL3-78B ↓1 64.8%
- 10
Qwen2.5-VL-72B ↓3 64.5%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The Multi-Crit paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Multi-Crit leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
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 Newest score check. The newest came 13 months ago, in October 2025, for Qwen3-VL-8B-Instruct. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (25 entries).
- Pass Models scored 17 models we track, plus 8 other entries on its leaderboard.
- Warning Newest score The newest came 13 months ago, in October 2025, for Qwen3-VL-8B-Instruct.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 (#71).
- Pass Headroom The best is 43%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on multi-crit.github.io: 17 rows for models we track and 8 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.