Mystery Model Ox-Alpha Scores 80%+ on SWE Benchmarks, Suspected GLM Release

A stealth model named Ox-Alpha reportedly scored about 80% on DeepSWE's 10-task test, beating GPT-5.6-sol (52%) and Fable (65%). Based on speed, TTFT, and cache hit patterns, users speculate it is a new Zhipu GLM release rather than DeepSeek.

2026-08-21 ~ 2026-08-21 · 3 related posts

• Mystery Model Ox Alpha Suspected to be Zhipu GLM — teortaxesTex · 2026-08-21
• Stealth Model 'Ox Alpha' Hits 80% on DeepSWE Benchmark — AccBalanced · 2026-08-21
• Stealth Model Ox-Alpha Released, Outperforms Fable on SWE Benchmark — troll_khan · 2026-08-21