OpenMLE: Open-Source AI4AI System Hits 71% on MLE-Bench Lite Using a Single RTX 4090
TianbaoX · x · 2026-07-31
FrontisAI launched OpenRSI, an open research series dedicated to recursive self-improvement (RSI), debuting with the full-stack AI4AI system OpenMLE. It uses evolutionary agents to improve ML solutions via executable feedback.
The system features three components: OpenMLE-Gym (5,758 executable tasks), OpenMLE-ERL (execution-grounded SFT + RL), and OpenMLE-Evo (experience-guided long-horizon search).
Performance: The trained Frontis-MA1-35B model paired with OpenMLE-Evo-Max achieves 71.21% Medal Average on MLE-Bench Lite, beating GPT-5.5 + Codex (68.18%) and closely trailing the top-tier Kimi K3 + Claude Code combo (72.73%). It is highly cost-effective, running on a single RTX 4090 (capped at 12GB VRAM) with a 12-hour per-task budget. The components also successfully transfer to 10 held-out NatureBench Lite tasks.
Related event: FrontisAI Open-Sources OpenMLE System and 35B Model(2 posts)→
More from coding & agent
- Connected steps are not a workflow: nobody designs the handoff — usehive · 2026-10-16
- Making Blender animations with Opus, Astra and Argon: how you prompt still matters — chongdashu · 2026-10-16
- Running PolyStrata + Cline coding agent on a $0-fancy laptop at ~90 tok/s — rpwoerk · 2026-10-16
- How do you attribute agent gains: model training or infrastructure? — jrxna · 2026-10-16
- Feed Claude your books, blogs and top engineers' code for claude.md — 'it now writes better code than me' — blaizedsouza · 2026-10-16
- Open-source /label plugin auto-names and colors Claude Code sessions in one step — hermanschutte · 2026-10-16