$1,000-trained HRM-Text shows Sapient bet on recurrence before OpenAI's Astra
rohanpaul_ai · x · 2026-09-06
- Sapient introduced its Hierarchical Reasoning Model (HRM) in June 2025 and open-sourced HRM-Text in May.
- The core idea: reasoning can spend compute by repeatedly updating internal state via recurrent loops, rather than relying only on a conventional transformer stack — the same direction OpenAI's GPT-6 Astra is now pushing into the frontier conversation.
- HRM-Text is a 1B-parameter open-source model pretrained from scratch on just 40B unique tokens for roughly $1,000 in GPU compute, with a full pretraining framework released for others to inspect, retrain, and build on.
Related event: Sapient's Open-Source HRM-Text Model Tops 250K Downloads(2 posts)→
More from Models
- Meta researcher: horrifying infra, not talent, is why OSS can't catch closed frontier — A_K_Nain · 2026-10-23
- TypeSafe's Jev: a "Decision Model" LLM That Returns Floats at $0.042/M Input Tokens — ariG23498 · 2026-10-23
- Microsoft's MAI-Base-2 scaling ladder pegs the L42 run at ~1.8×10^23 FLOPs — teortaxesTex · 2026-10-23
- Local model agentic coding benchmark: 27B matches 125B, tool-calling is the lever — Joralk_ · 2026-10-23
- NVIDIA open-sources PersonaPlex-7B, a full-duplex voice AI model that listens and talks simultaneously — solyarisoftware · 2026-10-23
- NVIDIA's KumoRFM-2: a foundation model that predicts churn, fraud and demand straight from relational databases — solyarisoftware · 2026-10-23