Cloudflare Workers AI Model Picker

Cloudflare Workers AI is a serverless AI inference service that runs machine learning models on Cloudflare's global edge network. You can call AI models directly from Cloudflare Workers, Pages Functions, or via the REST API — with no GPU setup, no model downloads, and ultra-low latency since models run close to your users.

Workers AI has a free tier: 10,000 neurons per day for most models. Beyond the free tier, you pay per neuron (a unit of compute). Pricing varies by model — larger models consume more neurons per token. Check the Cloudflare dashboard for current pricing. The free tier is generous enough for prototyping and low-traffic apps.

Workers AI offers: text generation (LLMs like Llama, Mistral, Gemma, Qwen), text embeddings (for vector search / RAG), image generation (Stable Diffusion), image classification, speech-to-text (Whisper), translation (M2M100), and more. New models are added regularly.

Add the AI binding in wrangler.toml: [ai] binding = 'AI'. Then in your Worker: const response = await env.AI.run('@cf/meta/llama-3.1-8b-instruct', { prompt: 'Hello' }). The binding handles authentication, model loading, and billing automatically.

AI Gateway is a layer in front of AI providers (including Workers AI, OpenAI, Anthropic, etc.) that adds caching, rate limiting, analytics, and cost tracking. Route your AI requests through AI Gateway to get observability, reduce costs with response caching, and protect against overuse.

Yes. Workers AI has a REST API compatible with the OpenAI API format. You can call it from any HTTP client using your Cloudflare account ID and API token. This is useful for testing models or integrating with non-Workers environments.