Create each key on its own page, paste it below, then copy the finished lines into your project's .env file. What you paste stays in this tab. It is never sent to Wirestack or saved.
Cloudflare Workers AI Account IDCLOUDFLARE_ACCOUNT_IDGet key
Click Use REST API. Your Account ID is shown there
Cloudflare Workers AI API tokenCLOUDFLARE_API_TOKENGet key
Click Use REST API, then Create a Workers AI API Token
On November 12, 2026 Cloudflare added Inkling 2 Small from Thinking Machines Lab to Workers AI. The model ID is @cf/thinkingmachines/inkling-2-small.
Input is text and images; output is text.
It supports function calling across several turns.
The context window is 1,048,576 tokens.
Reasoning is on by default at high. Set reasoning_effort to max, medium, low or minimal, or to none to turn it off. The model page also accepts auto (same as high) and xhigh (same as max).
Price: $0.30 per million input tokens, $1.00 per million output tokens, $0.10 per million cached input tokens.
Rate limit: 20 requests per minute on standard Workers AI billing, 50 with prepaid AI Gateway credits.
The model is not on the Workers Free plan. It needs the Workers Paid plan or prepaid AI Gateway credits. You can call it through the Workers AI binding (env.AI.run()), the REST API, the OpenAI-compatible endpoint or AI Gateway.
What to do
Nothing changes in an existing integration. To try the model, swap the model ID in a chat call. You need a Cloudflare account ID in CLOUDFLARE_ACCOUNT_ID and an API token in CLOUDFLARE_API_TOKEN (Cloudflare's sample calls it CLOUDFLARE_AUTH_TOKEN), on a paid plan. Cloudflare's REST example:
Shell
curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run/@cf/thinkingmachines/inkling-2-small \ -X POST \ -H "Authorization: Bearer $CLOUDFLARE_AUTH_TOKEN" \ -d '{ "messages": [{ "role": "system", "content": "You are a friendly assistant" }, { "role": "user", "content": "Why is pizza so good" }]}'
Reasoning runs at high unless you lower it, so set reasoning_effort for simple calls. Plan around the 20 requests per minute limit before you put the model behind user traffic.
Not confirmed
The announcement does not say how reasoning tokens are billed, or whether the rate limits can be raised on request.