Radar Telegram alerts Open Radar

Hugging Face and Llama.cpp hit Llama for Windows in Microsoft's keynote.

תרגם לעברית

Метаверсище и ИИще · · Source

Hugging Face and Llama.cpp hit Llama for Windows in Microsoft's keynote.

Lama for Windows is a lightweight Windows 11 system tray gadget that was announced as part of the Microsoft conference keynote. NET Conf. The utility, weighing only 1 MB, launches local neural networks with one click without console, manual environment configuration, and that's it.

The application functions as a local inference server with full compatibility with OpenAI and Anthropic APIs (port localhost:9931), and allows connecting coding agents, IDEs and extensions without API keys in the cloud. The engine automatically scans the computer configuration and selects the optimal weight quantization level, context limit, batch size, and decoding mode for the amount of available RAM and video memory. Well, I mean, it turns out he's kind of a token auditor.

All weights are stored in a single copy in the global cache of hugging faces and shared with other tools, and when inactive for more than 5 minutes, the model is automatically unloaded from memory to save system resources. In addition to the classic chat web interface, installation is available with a single command in PowerShell, and a release will appear in the official Microsoft store in the near future.

More on Radar

Want to see every report in real time?

Hundreds of sources worldwide, translated into your language, with alerts for what matters to you.

Sign up free
Share on WhatsApp Get alerts on Telegram