Surface Evolver Bench

What does it measure?
Tests whether an AI agent can use Surface Evolver, a program that models liquid surfaces shaped by surface tension.
Who evaluates it, and how?
The agent looks things up in the documentation and writes a complete input file meeting the given conditions (geometry, contact angles, gravity, volume). It is scored as the average partial credit across tasks.
Reading the score
Scores run from 0 to 100%. Higher is better.
Top score95.0%
Models tested25
Scoring basisHigher is better
Model release date

Can it run scientific simulation software to spec?

Surface Evolver Bench

  1. 🥇AnthropicClaude Fable 5Reasoning effort: High95.0%
  2. 🥇Moonshot AIKimi K395.0%
  3. 🥉OpenAIGPT-5.6 SolReasoning effort: Extra High93.1%
  4. #4OpenAIGPT-5.5Reasoning effort: High88.1%
  5. #5AnthropicClaude Opus 4.8Reasoning effort: High87.5%
  6. #6OpenAIGPT-5.6 TerraReasoning effort: Extra High83.8%
  7. #7GoogleGemini 3.8 FlashReasoning effort: High76.9%
  8. #8SpaceXSIGrok 4.5Reasoning effort: High74.4%
  9. #9OpenAIGPT-5.6 LunaReasoning effort: Medium61.9%
  10. #10AnthropicClaude Sonnet 5Reasoning effort: Medium60.0%
  11. #11GoogleGemini 3.5 FlashReasoning effort: Medium58.1%
  12. #12Z.aiGLM 5.2Reasoning effort: High55.6%
  13. #13MiniMaxMiniMax M355.0%
  14. #14MetaMuse Spark 1.1Reasoning effort: High52.5%
  15. #14Z.aiGLM 5.3 FlashReasoning effort: Max52.5%
  16. #16Moonshot AIKimi K2.7 Code48.8%
  17. #17DeepSeekDeepSeek V4.1 FlashReasoning effort: High46.3%
  18. #18AlibabaQwen3.6 35B A3B44.4%
  19. #19DeepSeekDeepSeek V4 ProReasoning effort: High40.0%
  20. #20GoogleGemma 4 31B30.6%
  21. #21Mistral AIMistral Large 428.1%
  22. #22Mistral AIMistral Medium 3.526.9%
  23. #23OpenAIGPT OSS 120B25.0%
  24. #24PoolsideLaguna M.115.6%
  25. #24Arcee AITrinity Large Thinking15.6%
Half of models ≤ 55.0%

The "harness" is the agent program the AI used to carry out the task. The same model can score very differently depending on its harness and reasoning effort.