GPT-6 'Bel' pretrain reportedly done at 10T+ params; GLM-5.3-Flash unmasked
WorldofAI · youtube · 2026-08-27
AI news roundup. Headline leak: OpenAI has reportedly finished a massive pretrain codenamed Bel, rumored at 10T+ total parameters, potentially the foundation for what comes after GPT-6.
Other items:
- Anthropic appears to be quietly testing Claude Marshmallow and Melon; speculation points to Fable 5.1 and another major Claude update soon
- Alibaba released Qwen3.8-Flash-Next open-weight preview: a 125B MoE with 6B active per token, a first look at the Qwen4 architecture with strong early coding results
- Zhipu's GLM-5.3-Flash is officially out, confirming the mysterious stealth model Ox Alpha was GLM all along: 320B total / 18B active params, MIT license
- Also covered: Gemini 3.5 Transcribe and OpenAI Jalapeño inference chip updates
Related event: Rumors: OpenAI Completes 10T-Parameter Pretraining of Model Codenamed Bel(12 posts)→
More from Models
- Omni 1.1 Flash model appears on Google Cloud Console — PleasantSir9581 · 2026-08-27
- Private benchmark: no major LLM scores above 80% on following instructions — crystalkalem · 2026-08-27
- Report: 1,200 OpenAI models talked to each other and schemed to hack their tests — Dan_Jeffries1 · 2026-08-27
- Chinese model progress driven by pretraining, not distillation, podcaster consensus argues — vista8 · 2026-08-27
- 8x RTX 3090 Setup Serves Qwen Flash Next at 661 tok/s with 262k Context — QuixiAI · 2026-08-27
- GLM-5.3 Flash Slammed: Unstable, 'Schizophrenic' Behavior — kevinnbass · 2026-08-27