Loading activity…
@miaai_lab
Building with AI & LLMs // Running frontier open models locally
AI
DeepSeek's README offers offline inference for DeepSeek-V4.1-Flash but lists additional systems needed for a complete service.
AI
DeepSeek says the model is live on its website, app and API, and optimized for DeepSeekHarness and Tencent CodeBuddy.
AI
A post interpreting the ministry update says high-risk open models may face closer scrutiny without ending support for open releases.
AI
ArtificialAnlys says that’s 1.8 times Super VL’s 46,000, but 28% fewer than Nemotron 3 Super’s 114,000.
Technology
Techmeme relays Anthropic researchers' prediction, naming smart ovens and off-brand Android TV boxes as examples.
AI
A user says they had Codex find the results, but cautions that they could be off.
AI
One user predicts that window from what they calculate as 52- to 83-day gaps between recent ZAI flagship releases.
Technology
Google says its new model has a 1 million-token output limit and is built for complex coding, enterprise knowledge work and cybersecurity defense. Broader access is planned after early testing.
AI
Google announces a 1 million-token output limit, up from 64,000, for its new model. Initial access goes to trusted cyber defenders through the Fairwind Program.
World & Business
A user links the price prediction to a belief that subsidized usage will end once Anthropic and OpenAI go public.
Technology
SpaceXAI says the update improves on Grok 4.6 at the same price and speed. TestingCatalog says its performance comes close to Opus 5 on certain tasks.
AI
A user describes a mixture-of-experts model with 27 billion active parameters, vision and a 1M context window, designed for coding, 3D and web design.
AI
Alibaba’s Qwen team says the 7-billion-parameter model can generate and edit transparent images and use up to 10 reference images. The publisher genztech.blog reports that its license is non-commercial only.
Technology
OpenAI says the framework sets criteria and timelines for public disclosure, including cases it hasn't fully explained or mitigated. More complex cases may require longer investigations.
AI
Qwen3.8-27B and a larger MoE model appear as open weights on Hugging Face.
AI
Open 30B MoE model with 3B active parameters targets always-on agents.
