Cmpnd Open-Sources dspy-wsp: Frontier Models Rewrite Small Model Weights for Free Gains

Cmpnd has open-sourced dspy-wsp, a DSPy optimizer that lets frontier models like GPT-6 directly rewrite small model weights for about $5-10, boosting decision models by up to 40 points without training. It supports MLX, PyTorch and vLLM, building on MIT CSAIL's Weight-Space Programming research.

2026-11-14 ~ 2026-11-14 · 3 related posts

• Cmpnd releases dspy-wsp: $5-10 of GPT-6 edits model weights, lifts every decision model tested — dbreunig · 2026-11-14
• dspy-wsp follow-up: MLX/PyTorch/vLLM backends, shines on decision models like d1-2B — dbreunig · 2026-11-14
• dspy-wsp open-sources weight-space programming: frontier LM edits model weights, no training, +40 pts on MATH-500 — dbreunig · 2026-11-14