Benchmark Almanac

MUSE

GPT-5.6 Sol 56.7% leads since September 2026
Not in model cards

Models answer questions about artworks used in language education. The tasks cover visual details, emotions, cultural context, and relationships within the image.

Made by
Luyao Zhu and colleagues
Released
September 2026
Kind
Other professions
Level
School
Scored
Checked automatically
Answers
Multiple choice
Measures
Capability
Language
English
In and out
Text, Images → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
A host
In views
Vision
Strongest model tested
GPT-5.6 Sol #14 on the capability index
Last model added
July 2026
Scores refreshed
November 20th, 2026
30
Entries in its paper, 29 models we track

GPT-5.6 Sol leads with 56.7% among models released by July 2026 GPT-5.6 Sol leads with 56.7% among models released by July 2026

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Table 2 ranks models by average performance. The headline is the mean of its twelve task scores. 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 July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-5.6 Sol 56.7%
  2. 2 Qwen3-VL-32B-Instruct 48.5%
  3. 3 Qwen2.5-VL-72B 43%
  4. 4 Qwen3-VL-8B-Instruct 42.1%
  5. 5 InternVL3-38B 41.6%
  6. 6 Qwen2.5-VL-32B 39.7%
  7. 7 InternVL3-14B 39.5%
  8. 8 GPT-4o (snapshot unspecified) 38.4%
  9. 9 Qwen2.5-VL-7B 35.6%
  10. 10 Gemma 3 27B 35.4%
  11. 11 Gemma 3 12B 35%
  12. 12 InternVL3-9B 33.9%
  13. 13 DeepSeek-VL2 33.1%
  14. 14 MiniCPM-V-2_6 33%
  15. 15 MiniCPM-o-2_6 32.9%
  16. 16 InternVL3-8B 32.4%
  17. 17 GLM-4V-9B 32.4%
  18. 18 MiniCPM-Llama3-V 2.5 32.1%
  19. 19 DeepSeek-VL2-Small 29.4%
  20. 20 Qwen2.5-VL-3B 29.2%
  21. 21 Gemma 3 4B 28.9%
  22. 22 llama3-llava-next-8b 28.6%
  23. 23 InternVL3-2B 28.3%
  24. 24 InternVL3-1B 27%
  25. 25 Yi-VL-6B 26.2%
  26. 26 CogVLM2-19B-Chat 25.8%
  27. 27 Yi-VL-34B 25.4%
  28. 28 LLaVA-v1.6-34B 22.6%
  29. 29 LLaVA-NeXT-72B 22.3%
  30. 30 DeepSeek-VL2-Tiny 22.2%

Source: The MUSE paper

#EntryScoreSettingDateSource
1 GPT-5.6 Sol 56.7% Jul 2026 Official leaderboard
2 Qwen3-VL-32B-Instruct 48.5% Oct 2025 Official leaderboard
3 Qwen2.5-VL-72B 43% Jan 2025 Official leaderboard
4 Qwen3-VL-8B-Instruct 42.1% Oct 2025 Official leaderboard
5 InternVL3-38B 41.6% Apr 2025 Official leaderboard
6 Qwen2.5-VL-32B 39.7% Mar 2025 Official leaderboard
7 InternVL3-14B 39.5% Apr 2025 Official leaderboard
8 GPT-4o (snapshot unspecified) 38.4% Official leaderboard
9 Qwen2.5-VL-7B 35.6% Jan 2025 Official leaderboard
10 Gemma 3 27B 35.4% Mar 2025 Official leaderboard
11 Gemma 3 12B 35% Mar 2025 Official leaderboard
12 InternVL3-9B 33.9% Apr 2025 Official leaderboard
13 DeepSeek-VL2 33.1% Dec 2024 Official leaderboard
14 MiniCPM-V-2_6 33% Aug 2024 Official leaderboard
15 MiniCPM-o-2_6 32.9% Jan 2025 Official leaderboard
16 InternVL3-8B 32.4% Apr 2025 Official leaderboard
17 GLM-4V-9B 32.4% Jun 2024 Official leaderboard
18 MiniCPM-Llama3-V 2.5 32.1% May 2024 Official leaderboard
19 DeepSeek-VL2-Small 29.4% Dec 2024 Official leaderboard
20 Qwen2.5-VL-3B 29.2% Jan 2025 Official leaderboard
21 Gemma 3 4B 28.9% Mar 2025 Official leaderboard
22 llama3-llava-next-8b 28.6% Apr 2024 Official leaderboard
23 InternVL3-2B 28.3% Apr 2025 Official leaderboard
24 InternVL3-1B 27% Apr 2025 Official leaderboard
25 Yi-VL-6B 26.2% Dec 2023 Official leaderboard
26 CogVLM2-19B-Chat 25.8% May 2024 Official leaderboard
27 Yi-VL-34B 25.4% Dec 2023 Official leaderboard
28 LLaVA-v1.6-34B 22.6% Jan 2024 Official leaderboard
29 LLaVA-NeXT-72B 22.3% May 2024 Official leaderboard
30 DeepSeek-VL2-Tiny 22.2% Dec 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.6 Sol 56.7% Jul 2026 Official leaderboard
2 Qwen3-VL-32B-Instruct 48.5% Oct 2025 Official leaderboard
3 Qwen2.5-VL-72B 43% Jan 2025 Official leaderboard
4 Qwen3-VL-8B-Instruct 42.1% Oct 2025 Official leaderboard
5 InternVL3-38B 41.6% Apr 2025 Official leaderboard
6 Qwen2.5-VL-32B 39.7% Mar 2025 Official leaderboard
7 InternVL3-14B 39.5% Apr 2025 Official leaderboard
8 Qwen2.5-VL-7B 35.6% Jan 2025 Official leaderboard
9 Gemma 3 27B 35.4% Mar 2025 Official leaderboard
10 Gemma 3 12B 35% Mar 2025 Official leaderboard
11 InternVL3-9B 33.9% Apr 2025 Official leaderboard
12 DeepSeek-VL2 33.1% Dec 2024 Official leaderboard
13 MiniCPM-V-2_6 33% Aug 2024 Official leaderboard
14 MiniCPM-o-2_6 32.9% Jan 2025 Official leaderboard
15 InternVL3-8B 32.4% Apr 2025 Official leaderboard
16 GLM-4V-9B 32.4% Jun 2024 Official leaderboard
17 MiniCPM-Llama3-V 2.5 32.1% May 2024 Official leaderboard
18 DeepSeek-VL2-Small 29.4% Dec 2024 Official leaderboard
19 Qwen2.5-VL-3B 29.2% Jan 2025 Official leaderboard
20 Gemma 3 4B 28.9% Mar 2025 Official leaderboard
21 llama3-llava-next-8b 28.6% Apr 2024 Official leaderboard
22 InternVL3-2B 28.3% Apr 2025 Official leaderboard
23 InternVL3-1B 27% Apr 2025 Official leaderboard
24 Yi-VL-6B 26.2% Dec 2023 Official leaderboard
25 CogVLM2-19B-Chat 25.8% May 2024 Official leaderboard
26 Yi-VL-34B 25.4% Dec 2023 Official leaderboard
27 LLaVA-v1.6-34B 22.6% Jan 2024 Official leaderboard
28 LLaVA-NeXT-72B 22.3% May 2024 Official leaderboard
29 DeepSeek-VL2-Tiny 22.2% Dec 2024 Official leaderboard

Rankings by part

We read 12 parts from the benchmark’s paper in 5 groups: visual perception, semantic understanding, affective interpretation, compositional reasoning and cultural. Each ranks models on one part of the test. GPT-5.6 Sol leads 7 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Object Classification

29 models we track have a score here.

  1. 1 GPT-5.6 Sol 76%
  2. 2 Qwen3-VL-32B-Instruct 54%
  3. 3 Qwen2.5-VL-72B 52%
  4. 4 InternVL3-38B ↑1 49%
  5. 5 InternVL3-14B ↑2 48%
  6. 6 Qwen3-VL-8B-Instruct ↓2 46.5%
  7. 7 Qwen2.5-VL-32B ↓1 44.5%
  8. 8 llama3-llava-next-8b ↑13 38%
  9. 9 MiniCPM-o-2_6 ↑5 34.5%
  10. 10 InternVL3-8B ↑5 33%

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 Sep 2026
GPT-5.6 Sol 56.7%

Source: The MUSE 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% 2024 2025 2026 Near the ceiling (85%) GPT-5.6 Sol Qwen3-VL-32B-Instruct Qwen2.5-VL-72B DeepSeek-VL2 GLM-4V-9B Yi-VL-6B 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5.6 Sol Qwen3-VL-32B-Instruct

Source: Official MUSE 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 Law, finance and medicine

All 637 in Law, finance and medicine
  • MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy. Qwen 3.8 Max 80.4%
  • HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
  • HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy. Ling-3.1-flash 81.7%
  • KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. o1-preview 92.7%
  • Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (30 entries).
  • Pass Models scored 29 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from July 2026, for GPT-5.6 Sol.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.6 Sol (#14).
  • Pass Headroom The best is 56.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 29 rows for models we track and 1 other entry 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