Thinking Machines Ships Inkling 2 Small, With Open Weights Planned Next Week
Thinking Machines Lab introduced Inkling 2 Small on Nov. 5: a 594B-parameter MoE with 17B active, image input and variable effort. They cite a 31 on…
etherninja.org Saving the Internet One Packet at a time!
Subscribe via RSS
Tag
Thinking Machines Lab introduced Inkling 2 Small on Nov. 5: a 594B-parameter MoE with 17B active, image input and variable effort. They cite a 31 on…
Ollama 0.40.15 is out. (Copilot CLI 1.0.106 was already covered in 2719, so this post is Ollama only.) Xiaomi MiMo’s mimo-v2.6 (9B, built for code and…
An update on Lemonade, the AMD-backed local AI server I covered in post 1527. Version 2026.45.3 is out, Phoronix reports. What’s new A fully redesigned desktop…
Revolut says it cut its AI coding inference costs by about 90% in under a year by making self-hosted models the default instead of commercial services…
Microsoft’s 4B Mage-VL-Cowork has me thinking about testing a small computer-use agent at home. This is a plan I’m considering, not something I’ve done. What I’m…
Gartner analysts Mark Petty and Raymond Paquet say “nine in 10 agent workflow steps are routine,” The Register reports, and open-weight models handle them “at parity…
Microsoft’s Mage team released Mage-VL-Cowork, a 4B “hybrid” computer-use agent. One vision-language policy can either click and type in a GUI from screenshots or drop to…
Ollama 0.40.14 is a fix release. ollama launch claude now applies the model’s output token limit for Claude Code, so cloud models no longer hit the…
Nace.AI released Drex Max, a dense 27B open-weight model built for decisions, not chat. You pass context and named questions, and it returns a probability for…
Today is Reformation Day, the day Luther posted his 95 Theses for debate. It got me thinking about writing down, in one public place, the principles…