Looking for Gemma4 like feature within Github Copilot to reduce the tokens usage

### Describe the feature or problem you'd like to solve

Gemma 4 like feature

### Proposed solution

Looking like Gemma feature, Please see below benefits that will help Github as well
0-Token Intent Classification
Every time you type a prompt, Gemini CLI must decide if it needs the "Big Brain" (Gemini Pro) or the "Fast Brain" (Gemini Flash).
- Without Gemma 3: The CLI sends your prompt to the cloud just to ask "Is this a hard question?" (Costs tokens/latency).
- With Gemma 3: It decides locally on your machine in milliseconds. Cost: 0 tokens.

2. Offloading "Boring" Tasks
Gemma 3 identifies tasks that don't need expensive AI:
- SQL Formatting: If you ask to format ERPCRP2.sql, Gemma routes this to a local formatter or a tiny model.
- File Reading: "List all columns in this CSV" is handled by the "Fast Brain" (Flash) because Gemma flagged it as simple.
- Result: You save your Gemini Pro quota for the hard stuff—like tracing that htimObjMgr log or debugging integration errors.

### Example prompts or workflows

_No response_

### Additional context

_No response_

The issue names no files, tests, or entry points to start from. First clarify the intended Gemma-like local classification and routing behavior; done would require a defined way to handle simple tasks locally while reducing cloud-token usage.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.