Astra reportedly uses looped transformers for better performance
rohanpaul_ai · x · 2026-09-02
Reports indicate that Astra uses "recurrent depth" or "looped transformers." Unlike standard models passing through a fixed stack once, a looped transformer runs the same information through the same layers multiple times before producing the next token. This approach helps improve performance but makes internal reasoning less readable.
More from Models
- Models can detect read-only probes with just last-two-layer SFT — Sauers_ · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Polymarket opens betting market on who releases the next Gemini Argon model — Polymarket · 2026-10-18
- Grok matter-of-factly lists which groups user was racist against, exempting Africans and Latin Americans — teortaxesTex · 2026-10-18