changelog
What's new in Qadenza.
Every beta build, newest first.
Mobile, MCP tools and thermal pause
- MCP tools. Give your models tools from a gallery, or connect any MCP server.
- Phones. New mobile layouts and gestures across the dashboard, and a refresh button when it's installed as an app.
- Several machines. Reworked server and model pages for organizations with more than one machine, and servers whose setup was skipped are flagged.
- Heat. A GPU at its thermal limit now pauses generation instead of ending the engine.
- Updates. Agents update themselves from a release channel.
- Password reset, chats that resume streaming after a reload, and cleaner model names.
The VS Code coding agent
- VS Code extension. A coding agent on your own model that reads your project and proposes edits.
- Multi-GPU control. Choose how llama.cpp splits a model across your GPUs.
- More models. Sharded GGUF downloads, and safetensors models converted to GGUF on the machine.
- Image generation on your own hardware.
- Fixes for llama.cpp models that hung while loading, and for generations left running after an agent restart.
Linux, multi-GPU and the relay
- Linux agent, with GPU discovery, runtime guidance and hardware metrics.
- Multi-GPU pooling, a CPU-only tier for machines without a GPU, and a disk-space check before downloads.
qadenza configto point an agent at your own relay, plus cleaner shutdown and uninstall.- Model manager showing Hugging Face model cards.
Your GPU, from anywhere.
Qadenza is in closed beta. Leave your email and we'll send an invite when a spot opens.
You're on the list. We'll email you when your invite is ready.