Surface Evolver Bench

Measures: 16 tasks where an agent writes Surface Evolver datafiles that model liquid surfaces under surface tension, gravity and volume constraints; graded with partial credit against hidden physical quantities.

Maintainer: Yann Henon
Introduced: 2026
Status: Active
Saturation: Saturated
Typical provenance: Independent
In scoring set: No

Official leaderboard

Leaderboard

Show independent results only

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

Leaderboard
Model	Mean score (partial credit)	Provenance	Source	Evaluated
Claude Fable 5 Anthropic · High effort	95.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Kimi K3 Moonshot AI · Unknown effort	95.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GPT-5.6 Sol OpenAI · Xhigh effort	93.1%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Kimi K3 Moonshot AI · Max effort	93.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GPT-5.5 OpenAI · High effort	88.1%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Claude Opus 4.8 Anthropic · High effort	87.5%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GPT-5.6 Terra OpenAI · Xhigh effort	83.8%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GPT-5.5 OpenAI · Medium effort	81.3%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Gemini 3.8 Flash Google DeepMind · High effort	76.9%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Grok 4.5 SpaceXAI (xAI) · High effort	74.4%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Claude Opus 4.8 Anthropic · None effort	68.1%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GPT-5.6 Luna OpenAI · Medium effort	61.9%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Claude Sonnet 5 Anthropic · Medium effort	60.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Gemini 3.5 Flash Google DeepMind · Medium effort	58.1%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GLM-5.2 Z.ai (Zhipu AI) · High effort	55.6%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
MiniMax-M3 MiniMax	55.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
GLM-5.3-Flash Z.ai (Zhipu AI) · Max effort	52.5%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Muse Spark 1.1 Meta · High effort	52.5%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Kimi K2.7 Code Moonshot AI	48.8%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
DeepSeek-V4.1-Flash DeepSeek · High effort	46.3%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Qwen3.8-27B Alibaba Group · Xhigh effort	45.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Qwen3.6-35B-A3B Alibaba Group	44.4%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
DeepSeek-V4-Pro DeepSeek · High effort	40.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Gemma 4 31B Google DeepMind	30.6%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Mistral Medium 3.5 Mistral AI	26.9%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
gpt-oss-120b OpenAI	25.0%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Laguna M.1 Poolside	15.6%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—
Trinity Large Thinking Arcee AI	15.6%	Third party	https://yhenon.github.io/surface-evolver-llm-eval/ via Epoch AI	—

Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.

Sources

Epoch AI