Inception Labs launches Mercury 3, its flagship diffusion LLM
Inception Labs says Mercury 3 is the most capable diffusion LLM released to date — more than 40 times larger than Mercury 1 in total parameters — claiming it outperforms GPT-6 Luna and Claude Haiku 5.5 while running four to eight times faster at over 1,000 tokens per second. The company is framing it as proof that text diffusion scales. We dug into the Mercury line's endpoints and pricing back in September; the flagship has moved fast since.
Hackers hit AI coding tool OpenCode, running up charges on users' accounts
OpenCode, the open-source AI coding agent that's become a popular alternative to Claude Code and Cursor, says hackers breached part of its online platform and stole users' API keys — then ran up charges on their AI accounts. Another reminder that agent tooling concentrates credential risk.
Nvidia launches new open models for enterprise agents and on-device PCs
Nvidia has released a new family of open models aimed squarely at two trends at once: enterprises wanting to run agents on their own infrastructure, and the push for private, on-device AI agents running on ordinary PCs.
Tenstorrent hits $6.2B valuation to scale open-weight AI chips
The Nvidia challenger raised $1.5 billion to remain independent after drawing acquisition interest earlier this year, and will put the money toward scaling its open-weight AI chip family.
🏠 Local models: MOSS-Image, FLUX.3 klein (ComfyUI build) and DeepSeek V4.1 Flash
Three worth a look this week. MOSS-Image — OpenMOSS's image model, repackaged by the ComfyUI team (Comfy-Org) as a ComfyUI-native single file under Apache 2.0 — has pulled over 670,000 downloads since October 16. FLUX.3 klein also shipped as a ready-to-run ComfyUI build from Black Forest Labs with more than 390,000 downloads; check the licence terms before commercial work. And DeepSeek V4.1 Flash landed with an MIT licence, FP8 builds and vision input — 1,652 likes in two days.