One year of Recursive Language Models: GPT-5-mini beats GPT-5 by 110% on 132k-token tasks, cheaper too
a1zhang · x · 2026-10-16
Alex Zhang marks one year since the first Recursive Language Models (RLM) blog post. RLM is an inference strategy where LLMs decompose and recursively interact with seemingly unbounded input prompts via a REPL environment. Key results: on OOLONG, RLMs with GPT-5-mini beat full GPT-5 by over 110% on 132k-token sequences at lower average query cost; on BrowseComp-Plus, RLMs with GPT-5 ingest 10M+ token "prompts" and answer highly compositional queries without degradation, even outperforming explicit indexing/retrieval. The takeaway: scaling frontier LLM context windows may be far easier than it sounds.
More from Models
- Nolla Health Post-Trains Nemotron 3.5 VL: Visual Diagnosis Jumps from 35% to 66.8% — LoganGrasby · 2026-10-16
- Dev argues coding models now beat nearly all engineers in every branch of software — zack_overflow · 2026-10-16
- Prediction: Chinese frontier models could match GPT-6 by March 2027 — teortaxesTex · 2026-10-16
- Liquid AI's d1-ultra, #1 on Artificial Analysis Decision Index, lands on Vercel — JosephJacks_ · 2026-10-16
- Weekly AI roundup: Claude Haiku 5.5 cuts prices ~90%, Gemini 4 Argon output jumps to 1M tokens — DeepLearningAI · 2026-10-16
- Tencent Hunyuan says reported agent activity came from third-party model evals, not deliberate targeting — TencentHunyuan · 2026-10-16