OmniPhys
Models solve Chinese physics problems from middle school through university and draw diagrams. The evaluator grades answers, reasoning, and generated figures.
- Made by
- East China Normal University and collaborators
- Released
- August 2026
- Kind
- Science questions
- Level
- University
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Subject
- Physical sciences
- Language
- Chinese
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision, Image, video and 3D generation
- Strongest model tested
- GPT-5.2 #51 on the capability index
- Last model added
- December 2025
- Scores refreshed
- November 20th, 2026
Gemini 3 Pro leads with 80.4 among models released by December 2025
Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in December 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
- 1
Gemini 3 Pro 80.4
- 2
Doubao-Seed-1.6 76.8
- 3 GLM-4.6V 71.1
- 4
Qwen3-VL-Plus 68.2
- 5 GPT-5.2 65.6
Source: The OmniPhys paper
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 80.4 | Nov 2025 | Official leaderboard | |
| 2 | | 76.8 | Jun 2025 | Official leaderboard | |
| 3 | GLM-4.6V | 71.1 | Dec 2025 | Official leaderboard | |
| 4 | | 68.2 | Sep 2025 | Official leaderboard | |
| 5 | GPT-5.2 | 65.6 | Dec 2025 | Official leaderboard |
Rankings by part
We read 21 parts from the benchmark’s paper, each ranking models on one part of the test, such as Test-Mini, Acoustics S1 and Acoustics S2. Gemini 3 Pro leads 15 of them, and Doubao-Seed-1.6 leads the other 6. Arrows show how far a model moved from its overall rank among the models we track.
Test-Mini
5 models we track have a score here.
- 1
Gemini 3 Pro 34.7
- 2
Doubao-Seed-1.6 22.4
- 3 GPT-5.2 ↑2 21.8
- 4 GLM-4.6V ↓1 19.9
- 5
Qwen3-VL-Plus ↓1 18.8
Acoustics S1
5 models we track have a score here.
- 1
Doubao-Seed-1.6 ↑1 79.8
- 2
Gemini 3 Pro ↓1 77.9
- 3 GLM-4.6V 70.8
- 4
Qwen3-VL-Plus 68.7
- 5 GPT-5.2 67.5
Acoustics S2
5 models we track have a score here.
- 1
Doubao-Seed-1.6 ↑1 84.6
- 2
Qwen3-VL-Plus ↑2 84.0
- 3 GPT-5.2 ↑2 81.7
- 4
Gemini 3 Pro ↓3 80.0
- 5 GLM-4.6V ↓2 79.0
Acoustics S3
5 models we track have a score here.
- 1
Doubao-Seed-1.6 ↑1 81.7
- 2
Gemini 3 Pro ↓1 78.9
- 3
Qwen3-VL-Plus ↑1 74.7
- 4 GLM-4.6V ↓1 73.4
- 5 GPT-5.2 67.8
Optics S1
5 models we track have a score here.
- 1
Gemini 3 Pro 79.8
- 2
Doubao-Seed-1.6 74.8
- 3 GLM-4.6V 69.2
- 4
Qwen3-VL-Plus 67.7
- 5 GPT-5.2 65.2
Optics S2
5 models we track have a score here.
- 1
Gemini 3 Pro 82.0
- 2
Doubao-Seed-1.6 81.0
- 3
Qwen3-VL-Plus ↑1 80.1
- 4 GPT-5.2 ↑1 79.7
- 5 GLM-4.6V ↓2 78.7
Optics S3
5 models we track have a score here.
- 1
Gemini 3 Pro 69.8
- 2 GPT-5.2 ↑3 64.4
- 3 GLM-4.6V 62.2
- 4
Doubao-Seed-1.6 ↓2 59.3
- 5
Qwen3-VL-Plus ↓1 52.9
Mechanics S1
5 models we track have a score here.
- 1
Gemini 3 Pro 83.6
- 2
Doubao-Seed-1.6 82.6
- 3 GLM-4.6V 75.0
- 4
Qwen3-VL-Plus 71.0
- 5 GPT-5.2 66.3
Mechanics S2
5 models we track have a score here.
- 1
Doubao-Seed-1.6 ↑1 86.5
- 2
Gemini 3 Pro ↓1 85.3
- 3
Qwen3-VL-Plus ↑1 83.0
- 4 GLM-4.6V ↓1 82.6
- 5 GPT-5.2 79.7
Mechanics S3
5 models we track have a score here.
- 1
Gemini 3 Pro 80.6
- 2 GPT-5.2 ↑3 72.3
- 3
Doubao-Seed-1.6 ↓1 72.0
- 4 GLM-4.6V ↓1 71.0
- 5
Qwen3-VL-Plus ↓1 68.3
Thermodynamics S1
5 models we track have a score here.
- 1
Gemini 3 Pro 86.5
- 2
Doubao-Seed-1.6 82.3
- 3 GLM-4.6V 76.3
- 4
Qwen3-VL-Plus 75.7
- 5 GPT-5.2 69.9
Thermodynamics S2
5 models we track have a score here.
- 1
Gemini 3 Pro 88.4
- 2
Doubao-Seed-1.6 87.2
- 3
Qwen3-VL-Plus ↑1 84.0
- 4 GLM-4.6V ↓1 83.2
- 5 GPT-5.2 82.1
Thermodynamics S3
5 models we track have a score here.
- 1
Gemini 3 Pro 81.5
- 2
Doubao-Seed-1.6 75.0
- 3 GPT-5.2 ↑2 73.8
- 4 GLM-4.6V ↓1 71.8
- 5
Qwen3-VL-Plus ↓1 67.9
Electricity S1
5 models we track have a score here.
- 1
Gemini 3 Pro 78.9
- 2
Doubao-Seed-1.6 75.2
- 3 GLM-4.6V 68.2
- 4
Qwen3-VL-Plus 65.5
- 5 GPT-5.2 60.1
Electricity S2
5 models we track have a score here.
- 1
Doubao-Seed-1.6 ↑1 81.8
- 2
Gemini 3 Pro ↓1 80.8
- 3
Qwen3-VL-Plus ↑1 79.0
- 4 GLM-4.6V ↓1 77.5
- 5 GPT-5.2 77.0
Electricity S3
5 models we track have a score here.
- 1
Gemini 3 Pro 73.7
- 2
Doubao-Seed-1.6 69.0
- 3 GPT-5.2 ↑2 66.3
- 4 GLM-4.6V ↓1 66.1
- 5
Qwen3-VL-Plus ↓1 65.9
Overall S1
5 models we track have a score here.
- 1
Gemini 3 Pro 81.7
- 2
Doubao-Seed-1.6 79.3
- 3 GLM-4.6V 72.1
- 4
Qwen3-VL-Plus 69.0
- 5 GPT-5.2 64.2
Overall S2
5 models we track have a score here.
- 1
Doubao-Seed-1.6 ↑1 84.4
- 2
Gemini 3 Pro ↓1 83.5
- 3
Qwen3-VL-Plus ↑1 81.4
- 4 GLM-4.6V ↓1 80.4
- 5 GPT-5.2 78.8
Overall S3
5 models we track have a score here.
- 1
Gemini 3 Pro 77.2
- 2
Doubao-Seed-1.6 70.2
- 3 GPT-5.2 ↑2 69.4
- 4 GLM-4.6V ↓1 68.5
- 5
Qwen3-VL-Plus ↓1 66.3
Overall Pobj
5 models we track have a score here.
- 1
Gemini 3 Pro 69.8
- 2
Doubao-Seed-1.6 61.4
- 3 GLM-4.6V 52.1
- 4
Qwen3-VL-Plus 45.6
- 5 GPT-5.2 39.4
Overall Popen
5 models we track have a score here.
- 1
Gemini 3 Pro 49.3
- 2 GLM-4.6V ↑1 33.1
- 3
Doubao-Seed-1.6 ↓1 32.8
- 4
Qwen3-VL-Plus 29.6
- 5 GPT-5.2 29.4
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The OmniPhys 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 OmniPhys 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 Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Gemini 4 Carbon 97%
- Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87%
- HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels.
Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 11 months ago, in December 2025, for GPT-5.2. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the benchmark’s paper (5 entries).
- Pass Models scored 5 models we track.
- Warning Newest score The newest came 11 months ago, in December 2025, for GPT-5.2.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 5 rows for models we track 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.