DeepSeek V4.1 Flash Launches at 70% Below V4 Pro: Pricing and Specs
DeepSeek V4.1 Flash cuts output pricing to $0.60 per million tokens off-peak — 70% below V4 Pro — with 1M context, 2,500 concurrency, and open weights on Hugging Face.
open weights
Nvidia Acquires Hugging Face for $12.93 Billion: What It Means for Open AI Models
Nvidia confirmed the $12.93B Hugging Face acquisition on Sept 3, 2026, with written openness commitments. What changes, what to watch, and how to insulate your projects.
State of Open Models Summer 2026: What Hugging Face's Numbers Actually Show
Hugging Face's Summer 2026 report: 2.96M open models, Chinese labs peaking at 2.78T parameters, AMD and NVIDIA top publishers, and 1.5% of repos taking 99.2% of downloads.
Nvidia Nemotron 3.5 Lightning: 30B Open Model for AI Agents With 4x Faster Output 2026
Nvidia launches Nemotron 3.5 Lightning, a 30B open-weights model built for AI agents, alongside the NeMo Switchyard routing library that cuts task cost to one-third.
Meta Muse Glimmer: نموذج 30 مليار معامل مفتوح المصدر للبرمجة والوكلاء الأذكياء — يعمل على جهازك 2026
Meta تطلق Muse Glimmer: نموذج 30B مفتوح الأوزان برخصة Apache 2.0 للتشغيل المحلي على بطاقة 24GB. وكيل ذكي للبرمجة واستخدام الأدوات واسترجاع الأخطاء — مجاني بالكامل.
Meta Muse Glimmer: 30B Open-Weights Agentic Coding Model That Runs on Your Device 2026
Meta releases Muse Glimmer: a 30B open-weights model under Apache 2.0 for local deployment on 24GB GPUs. An agentic model for coding, tool use, and failure recovery — completely free.
Shieldstral by Mistral: An Open-Source Safety Classifier Transforming AI Content Moderation
Mistral launches Shieldstral — a 3B open-weights safety model outperforming 7x larger competitors. Full analysis of the technology, performance, and applications.