Distilling frontier intelligence into small models.

3experiments
2open models
0GPUs
Method

Big AI teaches. Small AI learns.

Fine-tuning

The teacher writes the data.

Weight editing

The teacher rewrites the weights.

Benchmarks

Wins and losses, published.

Benchmarks

Every result. Even the misses.

Frontier agents train. Small models run offline.

AgentsClaude Code · Codex
First languageAzerbaijani
WeightsOpen