Elon: Grok 4.8 is a 2.5T parameter model, starts RL after training this week
ns123abc · x · 2026-09-14
Elon Musk says Grok 4.8 is a 2.5 trillion parameter model that will finish training this week on 'SpaceXAI's' new C++ stack before moving into RL. The claim is unverified — no official technical report yet — but if accurate it signals a major scale-up in both training infrastructure and post-training pipeline.
Related event: Musk Says 2.5T-Parameter Grok 4.8 Finishes Training This Week(2 posts)→
More from Infra
- New research: standard SGD matches AdamW for LLM RL training, with far less memory overhead — zhaoran_wang · 2026-09-14
- RTX 5090 stock nearly gone as prices spiral further — DustNearby2848 · 2026-09-14
- OneLA Shares Linear-Attention States Across Beams for 1.54-2.46x Faster Generative-Rec Decoding — _reachsumit · 2026-09-14
- UBTech Opens World's First 10,000-Unit Humanoid Robot Factory; SK's AI Data Center Hits 900MW — 创业邦 · 2026-09-14
- Deep dive: thermal gradients are HBM's #1 scaling bottleneck, says new architecture breakdown — blaizedsouza · 2026-09-14
- Perplexity launches Hybrid Compute to split AI tasks between cloud and local Mac — Aiden_Tech_Ai · 2026-09-14