Tag: Open Models
All the articles with the tag "Open Models".
-
China Is Making the Frontier Model Disposable
Alibaba and Zhipu released two cheap open-weight models in one day while OpenAI retired o3. The real power shift is not the leaderboard. It is making model loyalty look expensive.
-
AI Daily Digest - 2026-08-14
Open-weight model releases led the day, with Alibaba’s Qwen3.8-27B pushing a capable multimodal model into Apache 2.0 territory and broadening practical options for local and enterprise deployment. Meanwhile, Google and Cloudflare shipped infrastructure-level privacy and security tools that matter if you expect AI apps to move from demos into regulated, production environments.
-
AI Daily Digest - 2026-08-03
Two useful signals stood out today: Cloudflare detailed the inference engineering needed to run long-context MoE models economically, and ComfyUI shipped immediate support for MiniMax H3, an open-weights multimodal video model with native audio. Together, they show where the market is moving now: cheaper serving for demanding open models, and faster downstream tooling for local, creator-grade multimodal systems.
-
AI Weekly Digest - Week 31, 2026
This week’s clearest signal came from three layers of the stack at once: Moonshot positioned a new open frontier model, memory economics tightened again, and security teams got another reminder that models can uncover real zero-days. The broader pattern is practical, not flashy: shipping useful AI now depends as much on chips, trust, and developer tooling as on model size.