Physical-engineering benchmark: Claude Opus 5.5 tops at 48.3%, Grok 4.7 trails at 30.1%
teortaxesTex · x · 2026-10-24
A new benchmark, Physical-Engineering's Last Exam, tests models on real-world engineering tasks: Claude Opus 5.5 leads at 48.3%, followed by GPT 6 Astra (45.8%), Gemini 4 Argon (40.6%), and Grok 4.7 (30.1%), with paper, leaderboard, and code public.
teortaxesTex says the ranking is largely as expected and shares the full leaderboard: Sonnet 5.5 scores 43.5%, GLM 5.2 leads open models at 25.0%, DeepSeek-V4.1-Flash gets 14.4%, and Llama 4 Maverick sits at the bottom with 0.7%. He notes spatially loaded tasks are believed to favor models with high active parameter counts, and is curious how upcoming V4.1 and K3.1 will fare.
More from Models
- Terminal-Bench exposes NVIDIA Nemotron 3.5 near-zero score vs Kimi K3.1's 65.7% — JJitsev · 2026-10-24
- Bench shows GPT-5.6/6 Luna crush multi-agent coordination while DeepSeek V4 fails — JoshPurtell · 2026-10-24
- Grok Bot adds secure vault so agents can log in and pay without seeing your secrets — XFreeze · 2026-10-24
- Kimi K3.1 Pro reportedly first Chinese model to solve 20x20 nonograms — fairydreaming · 2026-10-24
- Kimi K3.1 vs GPT-6 Astra head-to-head on an identical agentic 3D task — GapNew4766 · 2026-10-24
- Kimi K3.1 Pro Launches: Scores 55 on Artificial Analysis, Open Weights Promised by Nov 2026 — Simon Willison · 2026-10-24