Mystery model ox-alpha likely Zhipu's GLM: token fingerprints match GLM-5.3 exactly
eyishazyer · x · 2026-08-21
An independent analyst ran forensic tests on the anonymous model ox-alpha instead of guessing: on a 2-second test video it consumed exactly 296 tokens — identical to GLM-5V-Turbo on the same clip across four separate videos. Tokenizing 25 prompts, ox-alpha's counts matched GLM-5.3 exactly, off only by a constant wrapper: same vocabulary, not just a similar model.
On the analyst's own 10-task DeepSWE subset (not an official leaderboard), ox-alpha scored 8/10 versus 65% or lower for Fable 5, GLM-5.3, Grok 4.6 and GPT-5.6-sol. Motive fits too: Zhipu's text-only GLM-5.3 shipped August 14 and was called out for lacking a vision flagship; weeks later a matching mystery model appeared. Pony Alpha previously pulled the same stealth drop before turning out to be GLM-5. The analyst's confidence is 90%; the free window ends around August 27, and prior stealth models from Zhipu, Xiaomi, Ant and Meituan were all claimed by their makers once revealed.
More from Models
- Muse Spark vs Qwen and GLM: US open models still trail on agentic tasks — zsakib_ · 2026-10-16
- Muse Spark 816B gets llama.cpp support just 30 minutes after release — ngxson · 2026-10-16
- Are the fastest GPU kernels now written by AI? Kernels are hyper-verifiable, argues one dev — fleetwood___ · 2026-10-16
- Scale AI launches muse spark open: 816B-param MoE with 1M context, billed as top US open-weight model — jiayi_pirate · 2026-10-16
- NVIDIA HF listing for Qwen3.8-Flash-Next-FP4-Local leaks unannounced Qwen3.8 model — TheZachMueller · 2026-10-16
- Muse Spark Open, billed as the most capable US open-weight model, lands on Nebius day one — Arindam_1729 · 2026-10-16