Leak: GLM-5 Was a Modest 40B-Active, 28.5T-Token Pretrain, Insider Says
teortaxesTex · x · 2026-08-22
Community poster teortaxesTex reacted to speculation that GLM 5.3 was delayed 2 months with a possible agentic-specialized "5.3 Turbo" variant, arguing it's not just about agentic specialization.
He claims (unverified) that GLM-5's pretraining was notably unambitious: off-the-shelf DSA, 40B active parameters, 28.5T tokens — mid-pack but good enough, and suggests Zhipu may have improved pretraining skills since.
More from Models
- Nathan Lambert slams Meta's Muse Open License as 'trash' — natolambert · 2026-10-16
- Google launches Gemini 4 Argon to Ultra and Pro subscribers, touting physics and 3D skills — ShikharMurty · 2026-10-16
- Google gives X users early access to test Argon, not just companies — lyraxana · 2026-10-16
- Grok Voice 3.0 Tops Design Arena Speech Leaderboard, Up 7.4 Points — XFreeze · 2026-10-16
- Robotics LLMs tested on 16 tasks: GPT-6 Astra hits 17.2% success vs Fable's 9.1% — pabbeel · 2026-10-16
- Meta open-sources Muse Spark: top US open model, scores just below DeepSeek — bindureddy · 2026-10-16