Benchmark Almanac

FrontierPhysics

Claude Opus 5.5 22.6% leads since Sep 2026 Not in model cards

56 specialist physics-research tasks in 9 domains. Graded by verifier pass plus a weighted rubric score over 3 trials.

v0.1. Each task ships a mentor package as a solvability control. README names benchflow-ai/FrontierPhysics, which returns 404. The public copy is under a contributor account.

Made by BenchFlow Released Aug 2026 Research science All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
wenjun-ke/FrontierPhysics 1 stars Leaderboard
19
Models scored on its own leaderboard

Claude Opus 5.5 leads with 22.6%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

19 models. Scroll sideways for the rest.

  1. 22.6 Claude Opus 5.5
  2. 17.9 GPT-6 Astra
  3. 16.1 GPT-6.1 Sol
  4. 13.1 Claude Fable 5.1
  5. 11.3 GPT-6 Sol
  6. 10.1 Claude Opus 5
  7. 8.30 GPT-5.6 Sol
  8. 8.30 Gemini 4 Argon
  9. 6.50 Grok 4.7
  10. 4.20 Claude Haiku 5.5
  11. 3.60 GPT-5.6 Luna
  12. 3.00 GPT-5.6 Terra
  13. 3.00 Grok 4.6
  14. 1.20 GLM-5.3
  15. 1.20 MiMo-V2.6 Pro
  16. 0.60 DeepSeek-V4-Pro
  17. 0.60 Kimi K3
  18. 0.60 Mistral Large 4
  19. 0.0 Gemini 3.8 Flash
  1. 1 Claude Opus 5.5 22.6%
  2. 2 GPT-6 Astra max 17.9%
  3. 3 GPT-6.1 Sol 16.1%
  4. 4 Claude Fable 5.1 13.1%
  5. 5 GPT-6 Sol 11.3%
  6. 6 Claude Opus 5 10.1%
  7. 7 GPT-5.6 Sol max 8.30%
  8. 8 Gemini 4 Argon 8.30%
  9. 9 Grok 4.7 6.50%
  10. 10 Claude Haiku 5.5 4.20%
  11. 11 GPT-5.6 Luna max 3.60%
  12. 12 GPT-5.6 Terra max 3.00%
  13. 13 Grok 4.6 3.00%
  14. 14 GLM-5.3 1.20%
  15. 15 MiMo-V2.6 Pro 1.20%
  16. 16 DeepSeek-V4-Pro 0.60%
  17. 17 Kimi K3 0.60%
  18. 18 Mistral Large 4 0.60%
  19. 19 Gemini 3.8 Flash 0%
#ModelScoreSettingDateSource
1 Claude Opus 5.5 22.6% Claude Code Sep 2026 Official leaderboard
2 GPT-6 Astra 17.9% max, Codex Sep 2026 Official leaderboard
3 GPT-6.1 Sol 16.1% Codex Sep 2026 Official leaderboard
4 Claude Fable 5.1 13.1% Claude Code Sep 2026 Official leaderboard
5 GPT-6 Sol 11.3% Codex Sep 2026 Official leaderboard
6 Claude Opus 5 10.1% Claude Code Jul 2026 Official leaderboard
7 GPT-5.6 Sol 8.30% max, Codex Jul 2026 Official leaderboard
8 Gemini 4 Argon 8.30% Antigravity Sep 2026 Official leaderboard
9 Grok 4.7 6.50% Grok Build Sep 2026 Official leaderboard
10 Claude Haiku 5.5 4.20% Claude Code Oct 2026 Official leaderboard
11 GPT-5.6 Luna 3.60% max, Codex Jul 2026 Official leaderboard
12 GPT-5.6 Terra 3.00% max, Codex Jul 2026 Official leaderboard
13 Grok 4.6 3.00% Grok Build Aug 2026 Official leaderboard
14 GLM-5.3 1.20% Claude Code Aug 2026 Official leaderboard
15 MiMo-V2.6 Pro 1.20% Claude Code Sep 2026 Official leaderboard
16 DeepSeek-V4-Pro 0.60% Claude Code Apr 2026 Official leaderboard
17 Kimi K3 0.60% Claude Code Jul 2026 Official leaderboard
18 Mistral Large 4 0.60% Claude Code Oct 2026 Official leaderboard
19 Gemini 3.8 Flash 0% Antigravity Sep 2026 Official leaderboard

More effort adds a median 3 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-6 Astra Codex
xhigh max
17.9%
GPT-5.6 Sol Codex
xhigh max
8.30%
GPT-5.6 Luna Codex
xhigh max
3.60%
GPT-5.6 Terra Codex
xhigh
3.00%
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 Claude Code 22.6%
GPT-6 Astra max Codex 17.9%
GPT-6.1 Sol Codex 16.1%
GPT-6 Astra xhigh Codex 14.9%
Claude Fable 5.1 Claude Code 13.1%
GPT-6 Sol Codex 11.3%
Claude Opus 5 Claude Code 10.1%
GPT-5.6 Sol max Codex 8.30%
Gemini 4 Argon Antigravity 8.30%
Grok 4.7 Grok Build 6.50%
Claude Haiku 5.5 Claude Code 4.20%
GPT-5.6 Luna max Codex 3.60%
GPT-5.6 Sol xhigh Codex 3.00%
GPT-5.6 Terra max Codex 3.00%
Grok 4.6 Grok Build 3.00%
GPT-5.6 Terra xhigh Codex 1.80%
GPT-5.6 Luna xhigh Codex 1.20%
GLM-5.3 Claude Code 1.20%
MiMo-V2.6 Pro Claude Code 1.20%
DeepSeek-V4-Pro Claude Code 0.60%
Kimi K3 Claude Code 0.60%
Mistral Large 4 Claude Code 0.60%
Gemini 3.8 Flash Antigravity 0%

4 models have held number one since its release in Aug 2026

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

Anthropic OpenAI
Own leaderboard 4 leaders
Claude Opus 5.5 22.6%

Scores over time

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% 2026 Near the ceiling (85%) Claude Opus 5.5 GPT-6 Astra Claude Fable 5.1 GPT-5.6 Sol DeepSeek-V4-Pro 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Science and reasoning

All 88 in Science and reasoning
  • GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Argon 96.4%
  • 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 Opus 5 90.1%
  • 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%