Benchmark Almanac

PHYSICS-Verified

GPT-6 Astra 91.1% leads since September 2026
Not in model cards

Corrected PHYSICS test problems with answer-level accuracy. The complete verified test set.

Original PHYSICS uses problem-level grading. Verified uses corrected problems and answer-level grading.

Released
September 2026
Kind
Science questions
Level
Expert
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
26
Entries on its own leaderboard, 19 models we track

GPT-6 Astra leads with 91.1% among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 12th, 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 newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

26 entries. Scroll sideways for the rest.

  1. 91.1 GPT-6 Astra
  2. 90.6 GPT-6 Sol
  3. 90.4 GPT-5.6 Sol
  4. 87.2 Qwen 3.8 27B
  5. 87.0 DeepSeek V4.1 Flash
  6. 85.3 GLM-5.3-Flash
  7. 82.6 GPT-6 Luna
  8. 80.7 Qwen3.5-122B-A10B
  9. 76.8 Gemma 4 31B IT
  10. 73.8 Qwen3.5-9B
  11. 61.7 Mistral-Small-4-119B-2603
  12. 34.4 Intern-S1-mini
  13. 31.7 Gemma 4 E4B
  14. 29.6 Ministral-3-14B-Instruct-2512
  15. 28.2 Llama-4-Scout-17B-16E-Instruct
  16. 27.7 Ministral 3 8B Instruct
  17. 24.6 InternVL3.5-38B
  18. 20.2 Gemma 4 E2B
  19. 19.4 Qwen3.5-2B
  20. 18.8 Ministral-3-8B-Reasoning-2512
  21. 16.4 InternVL2.5-38B
  22. 14.6 Phi-4-Reasoning-Vision-15B
  23. 9.70 Qwen2-VL-72B-Instruct-AWQ
  24. 9.20 Aria
  25. 5.40 Pixtral-12B-2409
  26. 3.00 DeepSeek-VL2-Small
  1. 1 GPT-6 Astra thinking 91.1%
  2. 2 GPT-6 Sol thinking 90.6%
  3. 3 GPT-5.6 Sol thinking 90.4%
  4. 4 Qwen 3.8 27B thinking 87.2%
  5. 5 DeepSeek V4.1 Flash thinking 87%
  6. 6 GLM-5.3-Flash thinking 85.3%
  7. 7 GPT-6 Luna thinking 82.6%
  8. 8 Qwen3.5-122B-A10B thinking 80.7%
  9. 9 Gemma 4 31B IT thinking 76.8%
  10. 10 Qwen3.5-9B thinking 73.8%
  11. 11 Mistral-Small-4-119B-2603 thinking 61.7%
  12. 12 Intern-S1-mini thinking 34.4%
  13. 13 Gemma 4 E4B thinking 31.7%
  14. 14 Ministral-3-14B-Instruct-2512 no thinking 29.6%
  15. 15 Llama-4-Scout-17B-16E-Instruct no thinking 28.2%
  16. 16 Ministral 3 8B Instruct no thinking 27.7%
  17. 17 InternVL3.5-38B no thinking 24.6%
  18. 18 Gemma 4 E2B thinking 20.2%
  19. 19 Qwen3.5-2B thinking 19.4%
  20. 20 Ministral-3-8B-Reasoning-2512 thinking 18.8%
  21. 21 InternVL2.5-38B no thinking 16.4%
  22. 22 Phi-4-Reasoning-Vision-15B thinking 14.6%
  23. 23 Qwen2-VL-72B-Instruct-AWQ no thinking 9.70%
  24. 24 Aria no thinking 9.20%
  25. 25 Pixtral-12B-2409 no thinking 5.40%
  26. 26 DeepSeek-VL2-Small no thinking 3.00%

Source: Official PHYSICS-Verified leaderboard

#EntryScoreSettingDateSource
1 GPT-6 Astra 91.1% thinking Sep 2026 Official leaderboard
2 GPT-6 Sol 90.6% thinking Sep 2026 Official leaderboard
3 GPT-5.6 Sol 90.4% thinking Jul 2026 Official leaderboard
4 Qwen 3.8 27B 87.2% thinking Aug 2026 Official leaderboard
5 DeepSeek V4.1 Flash 87% thinking Sep 2026 Official leaderboard
6 GLM-5.3-Flash 85.3% thinking Aug 2026 Official leaderboard
7 GPT-6 Luna 82.6% thinking Sep 2026 Official leaderboard
8 Qwen3.5-122B-A10B 80.7% thinking Feb 2026 Official leaderboard
9 Gemma 4 31B IT 76.8% thinking Apr 2026 Official leaderboard
10 Qwen3.5-9B 73.8% thinking Feb 2026 Official leaderboard
11 Mistral-Small-4-119B-2603 61.7% thinking Official leaderboard
12 Intern-S1-mini 34.4% thinking Official leaderboard
13 Gemma 4 E4B 31.7% thinking Mar 2026 Official leaderboard
14 Ministral-3-14B-Instruct-2512 29.6% no thinking Official leaderboard
15 Llama-4-Scout-17B-16E-Instruct 28.2% no thinking Official leaderboard
16 Ministral 3 8B Instruct 27.7% no thinking Dec 2025 Official leaderboard
17 InternVL3.5-38B 24.6% no thinking Aug 2025 Official leaderboard
18 Gemma 4 E2B 20.2% thinking Mar 2026 Official leaderboard
19 Qwen3.5-2B 19.4% thinking Feb 2026 Official leaderboard
20 Ministral-3-8B-Reasoning-2512 18.8% thinking Official leaderboard
21 InternVL2.5-38B 16.4% no thinking Dec 2024 Official leaderboard
22 Phi-4-Reasoning-Vision-15B 14.6% thinking Mar 2026 Official leaderboard
23 Qwen2-VL-72B-Instruct-AWQ 9.70% no thinking Official leaderboard
24 Aria 9.20% no thinking Sep 2024 Official leaderboard
25 Pixtral-12B-2409 5.40% no thinking Official leaderboard
26 DeepSeek-VL2-Small 3.00% no thinking Dec 2024 Official leaderboard

2 models have held number one on the official leaderboard since its release in September 2026

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

OpenAI
Own leaderboard Passed 85% in Sep 2026, by GPT-5.6 Sol
GPT-6 Astra 91.1%

Source: Official PHYSICS-Verified leaderboard

GPT-5.6 Sol on the official leaderboard passed 85% in July 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-6 Astra Qwen3.5-122B-A10B Ministral 3 8B Instruct InternVL2.5-38B Aria 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-6 Astra Qwen3.5-122B-A10B

Source: Official PHYSICS-Verified leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

OpenAI
PHYSICS-Verified Passed 85% in Sep 2026, by GPT-5.6 Sol
GPT-6 Astra 91.1%
PHYSICS Not tested since Mar 2025
o3-mini 59.9%
PHYSICS-Verified Passed 85% in Sep 2026, by GPT-5.6 Sol
GPT-6 Astra 91.1%

Source: Official leaderboards

Replaces PHYSICS

More in Science and reasoning

All 92 in Science and reasoning
  • GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Carbon 96.5%
  • 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%
  • 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%
  • 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%
  • 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%