[FEATURE] Automatic model routing based on task complexity (Cowork & Claude Code)

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Currently, users must manually select which Claude model to use (Opus, Sonnet, or Haiku) before starting a conversation. This creates a constant dilemma:

• If I pick Opus for everything, I'm wasting resources on simple tasks (overkill for basic questions or quick edits).
• - If I pick Sonnet or Haiku as my default, I miss out on Opus-level reasoning when I actually need it for complex tasks.
• - - By the time I realize I need a more powerful model, I've already lost context in the current conversation.

This is especially painful in Cowork, where non-technical users shouldn't have to think about model selection at all. Other AI platforms (ChatGPT, Gemini) already handle this automatically — Claude should too.

Related closed issues: #11549, #16620, #15721 — all requesting the same thing, showing strong community demand.

Proposed Solution

Add a "model": "auto" setting that intelligently routes each task to the right model:

• Haiku for simple questions, quick lookups, file reads, and routine edits
• - Sonnet for standard coding tasks, document creation, and multi-step workflows
• - - Opus for complex reasoning, architecture decisions, and deep analysis

The system should:

1. Analyze the complexity of each user query before selecting a model
2. 2. Show which model was selected (e.g., "Using Sonnet for this task") for transparency
3. 3. Allow users to override with a manual selection if needed
4. 4. Work across both Claude Code and Cowork

This should be the default experience in Cowork especially, where users are non-technical and shouldn't need to understand model differences.

Alternative Solutions

• opusplan only switches between plan/execute mode, not based on actual query complexity
• Manual /model switching breaks context and adds cognitive overhead
• Defaulting to Haiku/Sonnet sacrifices quality on complex tasks
• Competitors already solved this: ChatGPT auto-routes between GPT-4o and GPT-4o-mini, Google Gemini selects between Flash and Pro automatically. Claude is behind on this UX pattern.

Priority

High - Significant impact on productivity

Feature Category

API and model interactions

Use Case Example

1. I'm using Cowork to manage my daily workflow
2. I ask Claude "what time is it in Tokyo?" — this should use Haiku (simple lookup)
3. Then I say "create me a detailed financial report from this spreadsheet" — this should use Sonnet (document creation)
4. Then I ask "analyze this contract for potential legal risks and compare with industry standards" — this should use Opus (complex reasoning)
5. Currently I have to manually switch models or waste Opus tokens on simple questions. With auto-routing, I just work naturally and the system picks the right model.

Additional Context

Meta note: This entire issue — from researching prior requests (#11549, #16620, #15721), navigating GitHub, filling out this form, and writing every field — was done by Claude itself via Cowork's Chrome integration, at the user's request.

The fact that Claude can autonomously navigate GitHub, find related issues, and draft a feature request is a testament to how powerful Cowork already is. Now imagine if it could also auto-select the right model for each step of this process — that's exactly what we're asking for.

If Claude can do all this, it should also be smart enough to know when it needs Opus vs. when Haiku will do.