LangChain launches LangSmith Fine-Tuning with smithtune CLI to train models from agent traces
LangChain · x · 2026-09-25
LangChain released LangSmith Fine-Tuning (public beta) and the smithtune CLI, which converts agent trajectories stored in LangSmith into SFT datasets, trains via Fireworks or Baseten, and evaluates back in LangSmith — an end-to-end pipeline for building specialized models that can match frontier models on specific tasks at lower cost and latency. Teams can run it manually or let a coding agent drive the whole process.
Related event: LangChain Launches LangSmith Fine-Tuning to Train Models from Agent Traces(4 posts)→
More from coding & agent
- Ambion release gives agents a real workspace: SSH shell, SQL, and git repos — andreisavu · 2026-09-25
- Ambion runs Pi on AgentHarness, keeps model sessions per seat across exchanges — andreisavu · 2026-09-25
- Ambion 0.2.0 ships real workspace backends and modular npm packages — andreisavu · 2026-09-25
- The 'junior 10x developer' is a person plus constraints, not a person — l4rz · 2026-09-25
- The 'junior 10x developer' doesn't exist as a person, but as a person plus constraints — l4rz · 2026-09-25
- Have your agent wear a VM: VMs are orders of magnitude safer than containers — davidcrawshaw · 2026-09-25