Atlas

Benchmarks

← All benchmarks

Terminal-Bench Science — Engineering Sciences (Vals)

Science · 2026-09-23

Nine engineering-science tasks under Vals' Terminus 2 protocol with infrastructure/provider-error retries.

Top models (higher is better)

ModelScore
Claude Fable 5.133.3
Claude Opus 5.533.3
GPT-6 Astra33.3
Claude Opus 511.1
Gemini 3.8 Flash11.1
GLM-5.311.1
GPT-5.6 Luna11.1
GPT-5.6 Sol11.1
GPT-5.6 Terra11.1
Grok 4.611.1
Grok 4.711.1
Claude Fable 50.0
Opus 4.80.0
Sonnet 50.0
DeepSeek V4.1 Flash0.0
DeepSeek-V4-Flash-07310.0
DeepSeek V4 Pro 08130.0
Gemini 3.5 Flash0.0
Gemini 3.6 Flash0.0
Gemini 3.7 Flash0.0
GLM-5.3 Flash0.0
Hy4 Preview0.0
Kimi K30.0
Mercury 2.50.0
MiMo-V2.6-Flash0.0
MiMo-V2.6-Pro0.0
Muse Spark 1.30.0
Qwen3.8 27B0.0
Loading Atlas data…