Memdoor claims 5-model Qwen team beats flagship LLMs on LiveCodeBench Hard
PferdOne · reddit · 2026-11-01
A Reddit user shared Memdoor's open-source team-solve workflow: a team of five 27B-class Qwen models collaborating on coding tasks reportedly beats stronger flagship models on LiveCodeBench Hard. The claim is self-reported and unverified, but the multi-agent division-of-labor approach is open-sourced on GitHub.
More from coding & agent
- Mark Watson releases free Common Lisp book: 17 single-file programs covering LLMs and more — danbri · 2026-11-01
- Teleport Coding Challenge Enters Final Month: Agent Swarms Port NetHack, Fully Correct — davidbau · 2026-11-01
- How do you audit what an autonomous AI agent read, changed or shared? — whatthedevils1 · 2026-11-01
- Karpathy runs Opus on medium effort; reasoning-effort debates are the new tabs vs spaces — vincent_koc · 2026-11-01
- lattedit adds MCP so Claude Code and Codex can edit and compile LaTeX live — cneuralnetwork · 2026-11-01
- Thorsten Ball's newsletter: 8 parallel agents, Anthropic kills the coding interview — Register Spill (Thorsten Ball) · 2026-11-01