Ox Alpha Mystery Deepens: Speculation Points to a Post-Trained Chinese Model
brandon_galang · x · 2026-08-22
Speculation continues around the anonymous model Ox Alpha. @JacobRothfield ruled out Zhipu: GLM just released 5.3 and had to cut off its free-token hackathon for new users early; anonymous release might be a way to let the model be judged purely on performance.
brandongalang agrees and adds observations:
- Ox Alpha is clearly a Chinese model, or at least post-trained by a Chinese team — it answered his peptide research last night after Grok 4.6 refused, lacking typical Western safety guardrails.
- Given the absurd compute needed to serve the free token allowance with zero data retention, smaller Chinese labs can't support it.
- His guess: Composer 3 post-trained on a GLM 5.2 base — though this contradicts Cursor's claim that Composer 3 was trained from scratch.
Identity remains unconfirmed until the reveal.
Related event: Forensics Point to Zhipu's Unreleased GLM Behind Mystery Model Ox Alpha(31 posts)→
More from Models
- Unredacted Anthropic filing exposes internal data-quality debate months before Opus 3 cutoff — zacharynado · 2026-10-15
- Google says teams across the company are using Gemini 4 Argon on its hardest engineering problems — CurieuxExplorer · 2026-10-15
- Perplexity's new embedding models: index with a 9B model, search with 0.6B — antoine_chaffin · 2026-10-15
- 797GB of n-gram counts in CPU RAM let 1B models match 57% more compute — PangWeiKoh · 2026-10-15
- Leak: Moonshot's Kimi flagship 'x4' done with pre-training, K3.1 in internal testing — lyraxana · 2026-10-15
- Unsloth ships GGUF quants for Nemotron 3.5: Bolt-74B runs in 24GB RAM — danielhanchen · 2026-10-15