v2.24.1-glm52
A fork of Mistral Vibe v2.24.1 that adds the Z.ai GLM-5.2 model (hosted by Mistral) as a selectable option in the model picker. The default active model is unchanged.
What changed
A single entry was appended to DEFAULT_MODELS in vibe/core/config/vibe_schema.py so that zai-glm-5-2 shows up alongside the existing models in the picker:
- Model id / alias:
zai-glm-5-2(matches the Mistral API model id) - Thinking: off (the model card lists no reasoning mode)
- Images: not supported (text-only)
- Auto-compact threshold: 800,000 tokens (the model has a 1M context window)
- Pricing: as published on the model card (per million tokens)
Select it from the model picker inside the TUI, or pin it in your config:
active_model = "zai-glm-5-2"Model card
- Mistral docs: https://docs.mistral.ai/models/zai-glm-5-2
- Z.ai announcement: https://z.ai/blog/glm-5.2
Build from source
This is a source-only release — no prebuilt binaries are attached. GitHub auto-provides the tag source archive (.tar.gz and .zip) under "Assets". Build it yourself with uv:
# 1. Fetch the tag source
curl -LO https://github.com/simbo1905/mistral-vibe/archive/refs/tags/v2.24.1-glm52.tar.gz
tar xzf v2.24.1-glm52.tar.gz
cd mistral-vibe-v2.24.1-glm52
# 2. Install into an isolated venv (keeps any existing `vibe` install untouched)
uv venv ~/opt/mistral-vibe-fork --python 3.12
uv pip install --python ~/opt/mistral-vibe-fork/bin/python .
# 3. Run it from the absolute path, or symlink it onto your PATH
~/opt/mistral-vibe-fork/bin/vibeSet MISTRAL_API_KEY in your environment before launching.
Installing into a dedicated venv (rather than
uv tool install) lets this fork coexist with an official Mistral Vibe install — run the official one asvibeand this fork via its absolute path.
Disclaimer
This is an unofficial fork. The Z.ai GLM-5.2 model is a third-party open-weights model hosted by Mistral without Mistral modifications. This fork is not affiliated with or endorsed by Mistral AI or Z.ai.