Leaker Claims OpenAI's Next Big Pretraining Run Is Codenamed 'Bel'
ChrisGPT · x · 2026-08-26
ChrisGPT, an X account claiming to have first broken the story, says OpenAI's next major pretraining run is codenamed Bel — a "monster" model positioned as a fable killer, expected by year-end or a couple of months after Astra. The account says it had the codename six days before it circulated, but is unsure where the run sits in training and what post-training looks like, so declines to overstate the model's current state. Unverified — treat with caution.
Related event: Rumors: OpenAI completes 10T-parameter pretraining of model 'Bel'(13 posts)→
More from Models
- China's StepFun releases Step 5 Preview weights: 600B MoE, 27B active, 1M context — Dr_Singularity · 2026-10-15
- StepFun open-sources Step 5 Preview: 600B MoE with 1M context, No. 2 on OpenRouter — Dr_Singularity · 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