ccmr Update: Switch Among Six Chinese LLMs Like DeepSeek V4, Kimi K2.6, MiMo v2.5 Pro
The ccmr tool has been updated to support six major Chinese large‑language models—including DeepSeek V4, Kimi K2.6, and MiMo v2.5 Pro—offering a lightweight, isolated way to route Claude Code requests without affecting existing Anthropic subscriptions.
Recent Chinese LLM releases
In the last two weeks the following models were released: Xiaomi MiMo v2.5 Pro, DeepSeek V4 preview, Kimi K2.6, Alibaba Qwen 3.5 Plus, Zhipu GLM 5.1, and MiniMax M2.7.
Claude Code Model Router (ccmr)
ccmr (Claude Code Model Router) is a lightweight routing gateway created in December 2023. Its purpose is to let Claude Code users switch to domestic models without modifying the original Claude subscription or environment variables. Existing community solutions (e.g., ccr, cc‑switch) were considered too heavyweight.
Supported models after the latest update
DeepSeek V4 (upgrade from V3.2)
Qwen 3.5 Plus (unchanged)
Kimi K2.6 (upgrade from K2 Thinking)
GLM 5.1 (upgrade from 4.7)
MiniMax M2.7 (unchanged)
MiMo v2.5 Pro (newly added)
Testing of DeepSeek V4 Pro and MiMo v2.5 Pro showed that MiMo feels especially fast, consuming about 200 million tokens in half a day.
Installation and configuration
npm install -g claude-code-model-routerccmr initThe init command creates a .env file in the current directory. The user fills in the API keys for any subset of the six models; unfilled entries are ignored. ccmr start In a separate terminal run: ccmr claude Inside the Claude Code interface the following commands switch models instantly:
/model deepseek/model kimi/model glm/model qwen/model minimax/model mimoRuntime requirements
Even when using third‑party models, Claude Code still requires an Anthropic account for authentication.
Isolation from official Claude subscription
The routing flow is completely isolated from the official Claude subscription. The original Claude Code terminal continues to use the Opus/Sonnet subscription, while a separate ccmr terminal can be opened for cheaper or comparative runs. Both environments can operate simultaneously without interfering with each other.
Benchmark results
On the SWE‑Bench Verified benchmark, several of the domestic models achieve performance that matches or exceeds Claude Sonnet 4.5.
Recommendation
It is recommended to try the Chinese models with ccmr for cost‑effective experimentation.
Repository
GitHub: https://github.com/luwill/Claude-Code-Model-Router
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.