Focus ·
Agent Escape Enters the Postmortem Phase
A technical review of an OpenAI-related agent intrusion into Hugging Face is now circulating, while Modal says its platform isolation was not breached.
中文版17 linked articles
Open Models
Open-weight and open-source model releases, including Qwen, DeepSeek, Llama, and the ecosystems forming around them.
Latest intelligence
Focus ·
NVIDIA Reportedly Agrees to Buy Hugging Face for $12.9B
The Information reports that NVIDIA has agreed to acquire Hugging Face for $12.9 billion, potentially bringing a key open-model distribution layer into NVIDIA’s ecosystem.
中文版Focus ·
Qwen Opens Qwen3.8-27B for Agent Workflows
Qwen has released Apache 2.0 weights for Qwen3.8-27B, positioning the 27B local model for long-context coding and agent tasks.
中文版Focus ·
Hugging Face Reportedly Reaches $150M Annualized Revenue
The Information reports that Hugging Face’s annualized revenue has surpassed $150 million, up 50% in two months. Its enterprise subscriptions, private storage and model-distribution businesses are gaining relevance as AI infrastructure revenue streams.
中文版Focus ·
DeepSeek V4 Flash Becomes a Full Deployment Package
DeepSeek moved V4 Flash into API public beta with the 0731 model, new agent benchmarks, Codex support, published pricing, and MIT-licensed weights.
中文版Focus ·
Z.ai launches GLM-5.3 for cyber defense
Z.ai has released the open-source GLM-5.3 and positioned it for coding and cyber defense. The company says its cybersecurity capability is comparable to Anthropic’s Claude Mythos 5.
中文版Focus ·
Qwen opens weights for its 2.4T-parameter Max model
Qwen has released weights for Qwen3.8-2.4T-A95B, moving Qwen3.8-Max from limited hosted-preview channels to downloadable, self-deployable distribution.
中文版Focus ·
Meta Opens Muse Glimmer for Local Agents
Meta released the roughly 30B-parameter Muse Glimmer as open weights for local agent workflows, with a quantized model designed to fit systems with 24GB or 32GB of memory or VRAM.
中文版Focus ·
US Department of Energy introduces Genesis-Science-1
The US Department of Energy has launched the Genesis Open Models Initiative and introduced Genesis-Science-1 as its first open-weight model for scientific research.
中文版Focus ·
US AI testing may prioritize closed models
The White House has reportedly shared its cyber-testing framework with major AI labs without publishing it. Available details indicate that it may focus on closed frontier models with national-security risks and use secure government-access environments.
中文版Explore ·
Open-Source AI Security Coalition Gains Heat
NVIDIA and others are pushing the Open Secure AI Alliance as debate over open-source AI regulation and business risk continues to intensify.
中文版Focus ·
MiniMax Opens H3 Video Model Weights
MiniMax H3 weights are now downloadable from the official Hugging Face repository, moving the audio-video model from announcement to direct deployment and evaluation.
中文版Focus ·
Alibaba Moves Qwen3.8-Max Into Default API Access
Alibaba Cloud now lists qwen3.8-max as a hybrid-thinking model with thinking enabled by default, following a Token Plan-only preview.
中文版Focus ·
Kimi K3 Enters the Open-Weight Fight
Moonshot released weights for the 2.8 trillion-parameter Kimi K3, while the community is discussing its architecture and linear-attention variants.
中文版Explore ·
Agent Incident Reviews Could Become Required Procurement Reading
With a detailed Agent intrusion timeline and broader industry warnings, enterprise buyers may demand auditable incident boundaries and postmortems from suppliers.
中文版Explore ·
The First Barrier to Open Weights May Be 1.56TB
Although Kimi K3 releases 2.8-trillion-parameter weights, its 1.56TB size suggests open-source adoption may first be constrained by distribution and deployment burdens.
中文版Explore ·
Open-Model Competition May Shift Toward Architectural Legibility
High engagement with Kimi K3 architecture notes may indicate that open-source competition increasingly depends on whether communities can understand, dissect, and deploy a model.
中文版