AnalysisAI ModelsNovember 3, 2026

Qwen 3.8 27B on a laptop can interpret model weights, like GPT-6 Astra

Read original source →simonwillison.net

Simon Willison reproduced an example from the paper "Sparks of Weight-Space Programming" with Qwen 3.8 27B running locally, finding it could read raw floating-point weights after 45+ minutes. The paper reports GPT-6 Astra fixed Qwen3.5-0.8B's reasoning loops by editing 0.016% of weights, lifting MATH-500 from 7.9% to 48.1%.

2 sources

More stories today

Open the live feed