Yhn HealthMedicine first seen 1 d ago, last 58 min ago, peak #4

Memdoor team of five Qwen models beats Fable 5

Original: Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard

A project called Memdoor reports that a coordinated team of five Qwen3.8-27B language models outperformed Fable 5 on LiveCodeBench Hard, a demanding coding benchmark. The workflow, published on GitHub, shows multiple smaller models collaborating to solve problems that single larger models struggle with. The result is drawing attention because it suggests smaller open models, working together, can rival much bigger frontier systems.

Why now: People are discussing it because a team of small open-source models beating a leading model on a hard coding benchmark challenges assumptions about model size and performance.

MemdoorQwen3.8-27BFable 5LiveCodeBench

Open on hn →

Evidence

• Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard · guregodevo · 442

API: https://socialmediatrends-api.osmike.com/v1/trends/4425810