Chat, reason, call tools, and generate images through Cloudflare Workers AI. Explore the latest GLM, Kimi, Qwen, and DeepSeek models with a shared API key and usage dashboard.
Live usage
Estimated lifetime usage. Requests stop when the recorded total reaches the cap; concurrent requests can exceed it. This is separate from Cloudflare billing.
Spend over time
Spend by model
Catalog
Use a short alias or full model ID. Chat prices are USD per million tokens; image prices are per 1024 × 1024 output at the listed steps.
Checking catalog version…
| Model / provider | Capabilities / context | $ / 1M in | $ / 1M cached | $ / 1M out | Spent | Requests | |
|---|---|---|---|---|---|---|---|
| loading… | |||||||
Cloudflare platform notes
“Paid access” means Workers Paid or prepaid AI Gateway credits are required on the hosting account. A configured provider is not a guarantee of quota or model availability.
Workers AI includes 10,000 Neurons per day. Paid overage is $0.011 per 1,000 Neurons. This dashboard estimates model usage before account credits and free allocations.
Official pricing ↗Limits depend on the model and billing method. Cloudflare can return HTTP 429 for quota or capacity, and HTTP 403 when paid access is required.
Current limits ↗Older Llama 3 / 3.1 variants and Gemma 3 were retired on May 30, 2026. Kimi K2.5 moved to K2.6 with different pricing. Use explicit model IDs when migrating.
Deprecation details ↗Quickstart
Anywhere you can configure an OpenAI base_url + API key, you can use this. Pick your stack: