Atlas

Benchmarks

← All benchmarks

Physical-Engineering's Last Exam (hidden test, mechanical)

Professional Work · 2026-10-30

The 23 mechanical engineering tasks of Physical-Engineering's Last Exam's 56 hidden test tasks.

Benchmark source: Ple Bench

Top models (higher is better)

ModelScoreSource
GPT-6 Astra44.8Physical-Engineering's Last Exam: GPT 6 Astra on codex · Google
Claude Sonnet 5.543.2Physical-Engineering's Last Exam: Claude Sonnet 5.5 on agy · Google
Claude Opus 5.542.4Physical-Engineering's Last Exam: Claude Opus 5.5 on agy · Google
Claude Opus 540.5Physical-Engineering's Last Exam: Claude Opus 5 on agy · Google
Gemini 4 Argon36.1Physical-Engineering's Last Exam: Gemini 4 Argon on claude-code · Google
Claude Fable 533.9Physical-Engineering's Last Exam: Claude Fable 5 on opencode · Google
Gemini 3.8 Flash28.5Physical-Engineering's Last Exam: Gemini 3.8 Flash on claude-code · Google
Grok 4.727.3Physical-Engineering's Last Exam: Grok 4.7 on pi · Google
GLM-5.226.3Physical-Engineering's Last Exam: GLM 5.2 on pi · Google
Gemini 3.1 Pro Preview15.2Physical-Engineering's Last Exam: Gemini 3.1 Pro on openhands · Google
Qwen3-Coder-480B-A35B-Instruct6.9Physical-Engineering's Last Exam: Qwen3 Coder 480B on hermes · Google
Gemma 4 26B A4B IT6.3Physical-Engineering's Last Exam: Gemma 4 26B on openhands · Google
MiniMax M26.3Physical-Engineering's Last Exam: MiniMax M2 on hermes · Google
gpt-oss-120b5.7Physical-Engineering's Last Exam: gpt-oss 120B on codex · Google
Kimi K2 Thinking5.4Physical-Engineering's Last Exam: Kimi K2 Thinking on opencode · Google
Qwen3 Next 80B A3B Instruct4.4Physical-Engineering's Last Exam: Qwen3-Next 80B on openhands · Google
Llama 4 Maverick Instruct1.2Physical-Engineering's Last Exam: Llama 4 Maverick on pi · Google
Loading Atlas data…