# End-to-End Physics Research Benchmark for AI Agents (FrontierPhysics)

> FrontierPhysics, from BenchFlow, gives an AI agent a research project written by a working physicist and checks whether it gets the physics right and writes it up well enough to publish. It has 56 tasks across nine fields of physics, and each one is run three times.

Canonical page: https://benchlm.ai/benchmarks/frontierphysics

- Category: [Agentic](/agentic)
- Last updated: November 13, 2026

## About FrontierPhysics

- Year: 2026
- Tasks: 56 research projects from 42 physicists across nine fields, three trials each (168 rollouts per setup)
- Format: Agentic research in a sandbox: literature review, simulation or measurement, structured results and a written paper
- Difficulty: Open-ended research projects that took their authors about 270 days on average
- Paper: [FrontierPhysics: Evaluating Agents for End-to-End Physics Research (arXiv, November 4, 2026)](https://arxiv.org/abs/2611.06910)

Forty-two physicists from 26 institutions turned projects from their own research into tasks, and an admin team of physicists accepted 56 of 182 submissions. The tasks cover condensed matter, atomic and optical physics, quantum information, astrophysics, fluids, statistical physics, photonics, high energy and chemical physics, and mix theory, simulation and experiment; some give the agent a simulated lab device to measure. The agent reads a proposal, searches the literature, plans and runs the work, then hands in its numbers in a fixed format and a paper. A deterministic verifier checks the numbers against reference values, and a reviewer agent (GPT-5.6 Sol at xhigh effort in Codex) grades the paper, code and transcript against rubrics: global ones shared by every task, including checks for reward hacking and invented results, and task-specific ones the task author wrote. A rollout passes only if it passes every verifier test and every blocking rubric criterion. Each model runs in its lab's own agent tool where there is one (Claude Code, Codex, Antigravity or Grok Build); other labs' models run in Claude Code. We store the overall pass rate and keep the verifier pass rate, pass@3 and the paper's 0 to 100 Score (the rubric grade of rollouts that pass, zero otherwise) in each row note.

FrontierPhysics is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (21 models)

| Rank | Model | Creator | Score | Who ran it |
|------|-------|---------|-------|-----------|
| 1 | [Claude Opus 5.5](/models/claude-opus-5-5) | Anthropic | 27.4% | Run by FrontierPhysics |
| 2 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 23.2% | Run by FrontierPhysics |
| 3 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | OpenAI | 19.6% | Run by FrontierPhysics |
| 4 | [GPT-6.1 Terra](/models/gpt-6-1-terra) | OpenAI | 14.3% | Run by FrontierPhysics |
| 5 | [GPT-6 Sol](/models/gpt-6-sol) | OpenAI | 14.3% | Run by FrontierPhysics |
| 6 | [Claude Fable 5.1](/models/claude-fable-5-1) | Anthropic | 11.3% | Run by FrontierPhysics |
| 7 | [Claude Opus 5](/models/claude-opus-5) | Anthropic | 9.5% | Run by FrontierPhysics |
| 8 | [Gemini 4 Argon](/models/gemini-4-argon) | Google | 8.3% | Run by FrontierPhysics |
| 9 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenAI | 7.1% | Run by FrontierPhysics |
| 10 | [Kimi K3.1 Pro](/models/kimi-3-1-pro) | Moonshot AI | 6.5% | Run by FrontierPhysics |
| 11 | [Claude Haiku 5.5](/models/claude-haiku-5-5) | Anthropic | 4.2% | Run by FrontierPhysics |
| 12 | [GLM-5.3](/models/glm-5-3) | Z.AI | 4.2% | Run by FrontierPhysics |
| 13 | [Grok 4.7](/models/grok-4-7) | xAI | 4.2% | Run by FrontierPhysics |
| 14 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | OpenAI | 3.6% | Run by FrontierPhysics |
| 15 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | DeepSeek | 3.6% | Run by FrontierPhysics |
| 16 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | OpenAI | 3.6% | Run by FrontierPhysics |
| 17 | [Kimi K3](/models/kimi-k3) | Moonshot AI | 2.4% | Run by FrontierPhysics |
| 18 | [Grok 4.6](/models/grok-4-6) | xAI | 2.4% | Run by FrontierPhysics |
| 19 | [MiMo-V2.6-Pro](/models/mimo-v2-6-pro) | Xiaomi | 1.2% | Run by FrontierPhysics |
| 20 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | Google | 0.0% | Run by FrontierPhysics |
| 21 | [Mistral Large 4](/models/mistral-large-4) | Mistral | 0.0% | Run by FrontierPhysics |

## FAQ

### What does FrontierPhysics measure?

Whether an AI agent can carry a real physics research project from start to finish: read a proposal, search the literature, run simulations or measurements, report exact results and write a paper. A rollout counts only if its numbers match the reference and an AI reviewer passes the paper against the task author's rubric.

### Which model leads FrontierPhysics?

Claude Opus 5.5 in Claude Code leads at 27.4%, passing 46 of 168 rollouts. GPT-6 Astra at max effort in Codex is second at 23.2%, and GPT-6.1 Sol and GPT-6 Astra at xhigh effort tie at 19.6%; the paper finds these four cannot be separated statistically. The best model from a lab outside the US is Kimi K3.1 Pro at 6.5%.

### Why are FrontierPhysics scores so low?

Passing needs both correct numbers and research a physicist would accept. Agents often get the numbers but fail the rubric: Claude Opus 5.5 passes the verifier on 45.8% of rollouts but both checks on 27.4%. The paper finds the main weaknesses are loose reasoning and weak expert intuition, not missing physics knowledge or broken code.

## Compare Top Models on FrontierPhysics

- [Claude Opus 5.5 vs GPT-6 Astra](/compare/claude-opus-5-5-vs-gpt-6-astra)
- [GPT-6 Astra vs GPT-6.1 Sol](/compare/gpt-6-1-sol-vs-gpt-6-astra)
- [GPT-6.1 Sol vs GPT-6.1 Terra](/compare/gpt-6-1-sol-vs-gpt-6-1-terra)
- [GPT-6.1 Terra vs GPT-6 Sol](/compare/gpt-6-1-terra-vs-gpt-6-sol)