GPT-6 Astra edits Qwen3.5-0.8B weights directly: 0.016% of params changed, math up 6x
nathanborror · x · 2026-11-03
A team asked GPT-6 Astra to improve Qwen3.5-0.8B's math by editing its weights directly — no gradients, no training.
- Only 0.016% of weights were changed
- Fixed the small model's reasoning loops; math improved up to 6x, instruction following up to 3x
- They call it "Weight-Space Programming": frontier LLMs can natively read and write neural network weights
A notable glimpse of a new paradigm where large models directly optimize small-model weights.
Related event: MIT Paper Shows LLM Directly Editing Weights Boosts Small Model Math 6x(47 posts)→
More from Models
- Hands-on audit punctures Astra's '0.016% weight programming' claim: it just disabled Qwen's thinking mode — teortaxesTex · 2026-11-03
- GPT-6 edits Qwen3.5-0.8B weights directly: 0.016% of params changed, math up 6x — khademinori · 2026-11-03
- Leak: Unreleased astra6.2 builds a 3DS-style Mario Kart in 20 min under 1% quota — lxfater · 2026-11-03
- "Sonnet 5.5 used to be smarter": users sense capability regression — nlarusstone · 2026-11-03
- Apple Intelligence lands in China in December, powered by Alibaba's Qwen instead of ChatGPT — winer666 · 2026-11-03
- Only recent models seem to know "weight surgery" — are labs building Bayesian weight priors? — Tkaraletsos · 2026-11-03