Benchmark Almanac

Multi-Crit

o4-mini 43% leads since November 2025
Not in model cards

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.

Released
November 2025
Kind
General visual questions
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Images → Text
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
25
Entries in its paper, 17 models we track

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:

  1. 1 o4-mini 43%
  2. 2 o3 37.8%
  3. 3 GPT-5 37.5%
  4. 4 GPT-4o (snapshot unspecified) 36.4%
  5. 5 Gemini-2.5-Pro (snapshot unspecified) 35%
  6. 6 Claude 3.7 Sonnet 34.1%
  7. 7 InternVL3.5-38B 33.9%
  8. 8 MiMo-VL-7B 33.2%
  9. 9 Gemini 2.5 Flash Sep 2025 31.4%
  10. 10 Qwen2.5-VL-72B 30.5%
  11. 11 InternVL3-78B 29.2%
  12. 12 InternVL3.5-8B 28.8%
  13. 13 InternVL3-8B 26.5%
  14. 14 GLM-4.1V-9B-Thinking 23.6%
  15. 15 MiniCPM-V 4.5 22.1%
  16. 16 Eagle2.5-8B 19.7%
  17. 17 R1-Reward-7B 18.4%
  18. 18 LLaVA-Critic-R1-7B 17.5%
  19. 19 Qwen3-VL-8B-Instruct 17.5%
  20. 20 UnifiedReward-7B 15%
  21. 21 Molmo-7B 14.1%
  22. 22 Qwen2.5-VL-7B 13%
  23. 23 LLaVA-Critic-7B 12.6%
  24. 24 Llama 3.2 11B 11.5%
  25. 25 LLaVA-OneVision 7B 10.5%

Source: The Multi-Crit paper

#EntryScoreSettingDateSource
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 InternVL3.5-38B 33.9% Aug 2025 Official leaderboard
8 MiMo-VL-7B 33.2% Official leaderboard
9 Gemini 2.5 Flash Sep 2025 31.4% Sep 2025 Official leaderboard
10 Qwen2.5-VL-72B 30.5% Jan 2025 Official leaderboard
11 InternVL3-78B 29.2% Apr 2025 Official leaderboard
12 InternVL3.5-8B 28.8% Aug 2025 Official leaderboard
13 InternVL3-8B 26.5% Apr 2025 Official leaderboard
14 GLM-4.1V-9B-Thinking 23.6% Jul 2025 Official leaderboard
15 MiniCPM-V 4.5 22.1% Aug 2025 Official leaderboard
16 Eagle2.5-8B 19.7% Apr 2025 Official leaderboard
17 R1-Reward-7B 18.4% Official leaderboard
18 LLaVA-Critic-R1-7B 17.5% Official leaderboard
19 Qwen3-VL-8B-Instruct 17.5% Oct 2025 Official leaderboard
20 UnifiedReward-7B 15% Official leaderboard
21 Molmo-7B 14.1% Official leaderboard
22 Qwen2.5-VL-7B 13% Jan 2025 Official leaderboard
23 LLaVA-Critic-7B 12.6% Official leaderboard
24 Llama 3.2 11B 11.5% Sep 2024 Official leaderboard
25 LLaVA-OneVision 7B 10.5% Aug 2024 Official leaderboard
#ModelScoreSettingDateSource
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 InternVL3.5-38B 33.9% Aug 2025 Official leaderboard
6 Gemini 2.5 Flash Sep 2025 31.4% Sep 2025 Official leaderboard
7 Qwen2.5-VL-72B 30.5% Jan 2025 Official leaderboard
8 InternVL3-78B 29.2% Apr 2025 Official leaderboard
9 InternVL3.5-8B 28.8% Aug 2025 Official leaderboard
10 InternVL3-8B 26.5% Apr 2025 Official leaderboard
11 GLM-4.1V-9B-Thinking 23.6% Jul 2025 Official leaderboard
12 MiniCPM-V 4.5 22.1% Aug 2025 Official leaderboard
13 Eagle2.5-8B 19.7% Apr 2025 Official leaderboard
14 Qwen3-VL-8B-Instruct 17.5% Oct 2025 Official leaderboard
15 Qwen2.5-VL-7B 13% Jan 2025 Official leaderboard
16 Llama 3.2 11B 11.5% Sep 2024 Official leaderboard
17 LLaVA-OneVision 7B 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. 1 o4-mini 32.8%
  2. 2 Claude 3.7 Sonnet ↑2 31.8%
  3. 3 o3 ↓1 31.1%
  4. 4 InternVL3.5-38B ↑1 30.4%
  5. 5 GPT-5 ↓2 29.8%
  6. 6 InternVL3-78B ↑2 29.1%
  7. 7 Qwen2.5-VL-72B 28.4%
  8. 8 InternVL3-8B ↑2 26.1%
  9. 9 Gemini 2.5 Flash Sep 2025 ↓3 25.4%
  10. 10 InternVL3.5-8B ↓1 25.1%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Benchmark paper Not tested since Nov 2025
o4-mini 43%

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.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) o4-mini Claude 3.7 Sonnet Llama 3.2 11B 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) o4-mini Claude 3.7 Sonnet

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

All 636 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.

How we check