Inference provider · Inference API (serverless GPU)

Workers AI

Verified Aug 2026

Quick answer: Serverless open-model inference on Cloudflare’s edge GPUs — Workers binding or OpenAI-compatible REST, no cluster to run.

Best for

Serverless open-model inference on Cloudflare’s edge GPUs — Workers binding or OpenAI-compatible REST, no cluster to run.

Skip if

You need frontier Claude/GPT/Grok on one key (OpenRouter or Vercel AI Gateway), or you meant Cloudflare AI Gateway’s unified-billing layer.

Who it fits

  • Serverless open-model inference on Cloudflare’s edge GPUs — Workers binding or OpenAI-compatible REST, no cluster to run.
  • Builders who start in the browser without a local IDE setup

Pros

Serverless GPUs on Cloudflare’s edge — no cluster to provision; OpenAI-compatible REST plus Workers binding (`env.AI.run`); curated 50+ open-source models for text, image, embeddings, and more; pay-per-neuron with a free-plan allocation to start; same Cloudflare account as Workers, Vectorize, and R2.

Cons

Open-weight catalog only — not Claude, GPT, or Grok as first-party models; Cloudflare AI Gateway is a different product (logs, cache, optional unified billing); narrower lineup and limits than Together or Fireworks; best if you already live on Workers — not the default Cursor paste-key.

Facts

Locality:
Cloud-native
Surfaces:
Web
Maturity:
General availability
Open source:
Source available
Verified:
Aug 2026

Agent standards & memory

AGENTS.md: N/ASKILL.md: N/ARules / memory: Cloudflare API token + account id; or Workers AI binding.Free plan includes a neuron allocation; paid neurons after that.

Compare one neighbor first

Vendor pages sell hard. Read one alternative on this site, then leave for the official link if the fit still holds.

Alternatives

Related tools

Someone asks you about vibe-coding software? Share this site with them — we are adding more useful data regularly. agents.dancingteeth.net

← Back to full tool directory