MAI-Thinking

Reasoning playground

Hydrating page from local storage...

Run long-context reasoning prompts against Microsoft's MAI-Thinking-1 preview model.

Tools

Web grounding source

None Web Search WebIQ

Both use WebIQ credentials. Choose WebIQ for more verticals.

Additional tools

X MCP

Read-only access to X through MCP.

Connection

Foundry base endpoint: https://resource.services.ai.azure.com or https://resource.openai.azure.com

MAI-Thinking-1 calls the Foundry resource route /mai/v1/chat/completions. The deployment name is sent as model.

API

Foundry Resource Endpoint

Deployment Name

Authentication Mode

API Key

Stream

Advanced

Model controls

Max Completion Tokens

Use Reasoning Use Reasoning: Off

Reasoning Effort

Encrypted reasoning continuity Encrypted reasoning: Off

Enable to keep the opaque encrypted reasoning envelope in memory and return it unchanged on later turns.

Encrypted reasoning is returned on the final stream event or non-streaming response, kept only in memory, and passed back unchanged on later turns. It is redacted from payload previews, cURL, logs, auto-save, and exported sessions.

Template

System Prompt

Start a MAI-Thinking-1 conversation

Provide a reasoning task, configure your Foundry deployment, and generate an answer.