Skip to main content
Synth AI is the training and evaluation layer for LLM workflows. Use it to:
  • Run prompt optimization on fixed datasets (GEPA)
  • Run continual learning from live traffic and rewards (online MIPRO)
  • Run task apps locally (LocalAPI) or behind tunnels so the backend can evaluate them
If you’re looking for the open-source agent execution platform, see Horizons under the Horizons tab.

Prompt Optimization

Batch optimization (GEPA) on a dataset-backed task app

Continual Learning

Online improvement (MIPRO) from live rollouts and rewards