GitHub radar
Run the Xing4.0 AI model on your own computer, no server
Xing4.0 is an agentic AI model: it plans its own steps and uses tools. It used to run only on a server — now you can run it locally.
It's an AI agent — a model that decides its own steps and uses tools to get tasks done. It was built by China Telecom, a major Chinese phone carrier, not an AI lab.
When it helps
Good for local experiments and simple repetitive tasks without needing a server. For serious agentic coding work, stick with Opus or Fable in Claude Code — no Chinese model has matched them yet. The benchmark numbers come from the maker itself and aren't independently verified, so treat them with skepticism.
Pros
- Fast: only 4 of its 29 billion parameters run per answer
- Runs on a regular computer — no server needed
- Trained without any Nvidia GPUs, only on Huawei chips
- Context window up to 512,000 tokens of text
Cons
- Benchmarks are self-reported by the maker, not independently verified
- Weaker at serious coding than models like Opus and Fable
How to set it up — step by step
- 1Open your AI agent (Claude Code or similar) in your project folder
- 2Send the agent the text from the block below — it downloads and sets up the model
- 3Tell the agent how much memory your computer has (24 or 32 GB) so it picks the right file
- 4Check that the agent connected the model through Ollama or LM Studio
- 5Ask the agent to show you how to use the model in the terminal
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install the Xing4.0 model locally from this ready GGUF build: https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF - pick the version that fits my memory, set it up with Ollama or LM Studio, and show me how to work with it in the terminal.
You need a Mac or PC with at least 24 GB of memory - that fits the smallest version of the file (about 14 GB). 32 GB is safer - that fits the more accurate 19 GB version.
Open on Hugging Face▌ More finds