Provider pricing

Cloudflare Workers AI AI API pricing

Cloudflare Workers AI runs open models — Llama 4, Mistral, Qwen, DeepSeek, Stable Diffusion, Whisper — on Cloudflare's global GPU network. Pricing is per-neuron (Cloudflare's normalized compute unit) with a generous free tier and zero data-egress fees. Calls execute on the edge POP closest to the user.

Founded2023 (Workers AI launch — Cloudflare founded 2009)
HeadquartersSan Francisco, CA
Tracked modelsDirectory

Best for

Edge apps, serverless backends, and teams already on the Cloudflare stack.

Strengths

  • Inference at 300+ edge locations — sub-100ms TTFT worldwide
  • Per-neuron pricing with a generous always-free tier
  • Zero egress fees and tight integration with Workers and R2
  • AI Gateway built in for caching, rate-limiting, and observability