Xiaomi MiMo for Copilot Chat

Pick MiMo V2.5 Pro & V2.5 from the Copilot Chat model picker — with thinking mode, vision, and agent tools.

Love Xiaomi MiMo's reasoning capabilities but don't want to leave Copilot Chat? This extension drops MiMo V2.5 Pro & V2.5 straight into the model selector — with thinking mode, vision (V2.5), tool calling, and your own API key.

Why this extension?

• Don't replace Copilot — power it up. No new sidebar, no new chat UI to learn. Just new models in the picker you already use.
• Agent mode, tool calling, instructions, MCP, skills — all of it still works. Copilot's entire stack, now running on MiMo.
• Prompt caching that actually works. The extension sends prompt_cache_hit_tokens back to the API so MiMo can validate and continue its server-side cache — no warm-up waste, no surprise re-computation.
• Reasoning token tracking. Every response logs reasoning_effort feedback and exact reasoning token counts so you can see how much thinking the model invested.
• MiMo V2.5 supports vision. Drop screenshots, UI mockups, or diagrams into chat — V2.5 can see and understand them.
• MiMo V2.5 Pro excels at deep reasoning. Complex refactors, multi-step debugging, algorithm design — tasks that need serious thinking.
• BYOK, pay MiMo directly. Your API key, your bill, your rate limits. Stored in the OS keychain, never on disk.

Features

MiMo V2.5 Pro & V2.5 in the model picker

Both models show up alongside GPT-4o, Claude, and friends in Copilot Chat's model selector. 917K token context on both. Switch models mid-chat without losing history.

Prompt Caching with Full Feedback Loop

Most "compatible" extensions blindly forward API responses. This one closes the loop: it reads prompt_tokens_details.cached_tokens from each response and feeds it back to the API on the next request, ensuring MiMo's server-side prompt cache stays warm across multi-turn conversations. The result — dramatically lower costs and latency on long agent sessions.

Real-world cache performance (from actual usage logs):

[02:10:37.592] tokens: prompt=14973 completion=130 | cache: hit=12288 rate=82% | reasoning=83 | chars/tok=3.17
[02:10:43.595] tokens: prompt=15179 completion=230 | cache: hit=14912 rate=98% | reasoning=111 | chars/tok=2.58
[02:10:48.067] tokens: prompt=17110 completion=245 | cache: hit=15168 rate=89% | reasoning=42 | chars/tok=2.23
[02:10:53.388] tokens: prompt=17635 completion=174 | cache: hit=17088 rate=97% | reasoning=77 | chars/tok=1.99
[02:11:06.325] tokens: prompt=18068 completion=521 | cache: hit=17600 rate=97% | reasoning=108 | chars/tok=1.83
[02:13:50.253] tokens: prompt=19077 completion=57  | cache: hit=18048 rate=95% | reasoning=24 | chars/tok=1.75
[02:13:53.348] tokens: prompt=19201 completion=142 | cache: hit=19072 rate=99% | reasoning=23 | chars/tok=1.69

Cache hit rates climb to 97–99% on subsequent turns — meaning the model skips re-reading your entire conversation history and jumps straight to the new content.

Thinking Mode with Reasoning Token Visibility

Full support for MiMo's reasoning_content. Watch the model's thought process in real-time as it tackles complex problems. Every response also reports the exact number of reasoning tokens consumed — so you know how much "thinking" each answer cost you.

Vision Support (MiMo V2.5)

Drop a screenshot or UI mockup into chat and V2.5 will analyze it directly. Perfect for understanding code screenshots, design specs, and visual debugging.

Inherits Every Copilot Capability

Because this plugs into Copilot's native provider API, you get the full stack for free:

• Agent mode — autonomous multi-step tasks
• Tool calling — file edits, terminal, workspace search, Git, tests
• Instructions & skills — all your .instructions.md, AGENTS.md, and skills just work
• Prompt caching stats — MiMo's cache hit rate logged in the output channel so you can see the savings

Multi-Region API Support

Choose the API endpoint that matches your subscription plan:

• Default API (no plan)
• Token Plan — China
• Token Plan — Singapore
• Token Plan — Europe (Amsterdam)

Secure by Default

API key lives in VS Code's SecretStorage (OS keychain on macOS / Windows / Linux). Never in settings.json, never in your Git history.

Zero Runtime Dependencies

Pure VS Code API + Node.js built-ins. No Python, no Docker, no local proxy server to babysit.

Getting Started

Prerequisites

• VS Code 1.116 or later. This extension relies on non-public Copilot Chat APIs that may break on newer VS Code versions — report an issue if you hit one.
• GitHub Copilot subscription (Free / Pro / Enterprise — the free tier works)
• MiMo API key from platform.xiaomimimo.com

Usage

1. Install from the VS Code Marketplace
2. Run MiMo: Set API Key from the Command Palette (Ctrl+Shift+P)
3. Paste your API key
4. Open Copilot Chat, click the model picker, pick MiMo V2.5 Pro or MiMo V2.5
5. That's it — chat away

Models

Both support thinking mode, tool calling, and 917K token context.

Settings

Commands

License

MIT