DeLM project page: +17.5 points over strongest baseline with shared-context agents
rohanpaul_ai · x · 2026-10-08
The DeLM project page by Stanford researchers (Yuzhen Mao, Jerry Gu, et al.) details the mechanism: a shared context and task queue let agents coordinate asynchronously with no main agent. On Terminal-Bench 4.0, DeepSWE v1.1, and SWE-bench Verified, DeLM is faster and more accurate than Codex, Claude Code, their native subagents, and AOrchestra — up to 17.5 points more accurate than the strongest baseline and up to 2.49× faster than the harness it builds on. Back-end models: GPT-6-Astra (Codex) and Claude Opus 5.5 (Claude Code) at xhigh reasoning effort. Paper, code, traces, and plugin links included.
Related event: Stanford's DeLM Decentralized Multi-Agent Coding Runs 2.49x Faster(3 posts)→
More from coding & agent
- Codex plus a text-to-CAD plugin generates a radial aircraft engine — creatoroff · 2026-10-08
- Dev dumps 6x-Sol over quality, burns $200 Astra sub in half a day, moves to Claude Code — Late_Change5029 · 2026-10-08
- This Solo Dev Runs His Entire Business From One Obsidian Vault With AI — dSebastien · 2026-10-08
- Cognizant to Hire 1,500 US Graduates; Devin Cut Freight Firm's Rebuild Costs 37% — shashib · 2026-10-08
- Cisco Brings Claude Managed Agents to Webex With Its Own Governance Layer — shashib · 2026-10-08
- Dev finds 6.1 Sol surprisingly good at designing native iOS apps — Dimillian · 2026-10-08