GPT-6 Astra rumor: 1.2T active params, trained on Abilene's 100-150k GPUs
teortaxesTex · x · 2026-09-07
Discussion around rumored specs for GPT-6 Astra (codename Doug): 1.2T active parameters, 10T total. OpenAI confirmed the August 1 math result came from an internal Astra version; with Abilene handing over 100-150k GPUs after Spud's (GPT-5.5) pretraining ended in late March, the pretraining run is estimated at 90-110 days.
teortaxesTex is skeptical: if 1.2T active params were truly needed, that would signal inefficiency — he estimates <350B active params, >150T tokens, and >50% of compute spent in the RL stage, while noting current models still can't help implement basic research papers. All unconfirmed speculation.
Related event: GPT-6 Astra Rumors Swirl Over 1.2T Active Parameters(2 posts)→
More from Models
- Week 3 of "Fable 5.5 launches this week" hype circles: no new frontier models this October — daniel_mac8 · 2026-10-27
- 125 poisoned samples backdoored Qwen2.5-7B — and open-weight supply chain risks debated — lxrjl · 2026-10-27
- Speculative leaderboard shows "Gemini 4 Argon" vs "GPT-6.1 Sol"; observer calls MiniMax parity suspicious — teortaxesTex · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27
- Kimi K3.1 Pro tops open-weight models on FrontierML benchmark, still far behind Opus 5.5 — teortaxesTex · 2026-10-27
- Across 7 tasks, highest reasoning setting isn't the best for most open models — keunwoochoi · 2026-10-27