Skip to main content
Synth AI uses usage-based pricing. Generally speaking, we endeavor to provide core infrastructure (prompt optimization, graphs, and verifiers) with as little margin as possible. For now, we bill a direct passthrough but we do reserve the right to add enough margin to cover edge-cases and failures to keep the economics neutral at scale. We deeply believe in bringing more optimization and experimentation to AI builders, and driving down costs is a big part of that. Current usage-based surfaces include:
  • Prompt optimization (GEPA + MIPRO)
  • Graph optimization and inference (Graph Evolve / GraphGen)
  • Verifiers and RLM scoring
  • Environment Pools (managed sandboxes + browser pools)
  • Inference and evaluation jobs
Horizons is open source and can be self-hosted; managed hosting is handled separately.