Stanford's DeLM Decentralized Multi-Agent Coding Runs 2.49x Faster

Stanford's DeLM replaces the central coordinating agent with a shared context and task queue, letting agents asynchronously claim tasks, share findings, and correct each other. It achieves 2.49x faster multi-agent programming and 17.5 points higher accuracy over the strongest baseline, though evaluation setups vary.

2026-10-08 ~ 2026-10-08 · 3 related posts

• DeLM's decentralized multi-agent system runs 2.49x faster, but MAS evals pick wildly different metrics — jyangballin · 2026-10-08
• Stanford paper: decentralized multi-agent DeLM beats Claude Code and Codex, 2.49× faster — rohanpaul_ai · 2026-10-08
• DeLM project page: +17.5 points over strongest baseline with shared-context agents — rohanpaul_ai · 2026-10-08