Black-box fingerprint confirms Ox Alpha is based on GLM-5.3
FlunkyGraphics · reddit · 2026-08-21
Three black-box fingerprint tests confirm that Ox Alpha is based on the GLM-5.3 architecture:
- Tokenizer Consistency: Across 6 text types (EN/DE/CN/code/emoji), Ox Alpha's prompt tokens are exactly 75 higher than GLM-5.3 on every text. This indicates an identical tokenizer plus a constant 75-token hidden system prompt. Counts for Kimi/Qwen/MiMo/MiniMax diverge completely.
- Identical Error Strings: Invalid parameters trigger the exact same error message on Ox Alpha as on GLM 5.3: "[1210] This model always engages in thinking and cannot be disabled; please use low, high, or max".
- Temp-0 Output Similarity: Under greedy decoding, both produce identical Markdown quirks (e.g., German decimal comma 0{,}375) and near word-for-word matches on factual answers. Qwen/MiMo/Kimi format these completely differently.
Conclusion: Ox Alpha is highly likely a GLM model, possibly a vision variant of GLM 5.3 (GLM 5.3V) or an unannounced new version like GLM 5.5.
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
- 24 AI Systems Recommend $314 Per Visit in Care vs Physicians' $97, a 3.2x Gap — dvir_a · 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