OpenAI's Altman Unveils Astra as a New Step Toward AGI | The Close 9/3/2026
youtube.comBloomberg's The Close interview with OpenAI CEO Sam Altman discussing the GPT-6 Astra model, its capabilities, safety guardrails, cybersecurity risks, pricing strategy, and its role as a step toward AGI.
NVIDIA Nemotron 3 Ultra Powers Faster, More Efficient Reasoning for Long-Running Agents
developer.nvidia.comNVIDIA launches Nemotron 3 Ultra for efficient long-running agent reasoning, the Nemotron 3.5 Content Safety guardrail model, and moves releases to the OpenMDW-1.1 license.
China's Alibaba launches Qwen3.8-Max AI model with 2.4T parameters, 1M token context window
technode.globalAlibaba launched Qwen3.8-Max, its largest Qwen model with 2.4 trillion parameters and a 1M-token context window, ranking fifth in Text Arena and second in Vision Arena, with weights scheduled for release the following week.
Is Qwen Code Still Free? The 2026 Free-Tier Shutdown (and 3 Free Alternatives)
inventivehq.comAlibaba discontinued Qwen Code's free OAuth tier on April 15, 2026, cutting daily free requests from 1,000 to 100 and then closing the free hosted inference; the CLI remains free and open source with local/OpenRouter alternatives.
Alibaba Adds to China AI Breakthroughs With New Qwen Model
bloomberg.comBloomberg reports Alibaba released Qwen3.8-Max, built on 2.4 trillion parameters, claiming performance on par with or better than Anthropic's Fable 5 on several benchmarks, making it one of the largest AI models ever announced.
Bringing new Veo 3.1 updates into Flow to edit AI video
blog.googleGoogle rolled out Veo 3.1 and enhanced creative tools in Flow, bringing audio to capabilities like Ingredients to Video, Frames to Video, and Extend, with richer audio, more narrative control, and enhanced realism.
OpenAI pulls the plug on its Sora AI video app
cbsnews.comOpenAI said it is discontinuing its Sora AI video app and API, citing waning popularity and a shift of the research team toward world simulation for robotics; the move also ends its billion-dollar Disney deal. The app had released Sora 2 in September 2025 and sparked controversy over depictions and AI 'slop'.
Microsoft Phi & MAI · innFactory AI
innfactory.aiOverview of Microsoft's Phi open-weight family and the proprietary MAI family (MAI-Transcribe-2, MAI-Thinking-1, MAI-Code-1-Flash, MAI-Voice-2, MAI-Image-2.5) for edge and local deployments.
Amazon Nova AI models power business, productivity, and creativity
aboutamazon.comGuide to Amazon Nova foundation models including Nova 2 Sonic, multimodal embeddings and creative content generation models available on Amazon Bedrock.
With its latest Phi-4 reasoning model, Microsoft reckons bigger isn't always better - The New Stack
thenewstack.ioMicrosoft released Phi-4-Reasoning-Vision-15B, an open-weight multimodal reasoning model demonstrating that compact models trained on high-quality data can compete with much larger systems.
Microsoft Phi-4 in 2026: The Open-Weight Small Model That Runs on Your Laptop | FrankX
frankx.aiAnalysis of the Phi-4 family as of June 2026, covering the seven MIT-licensed variants from Phi-4-mini through Phi-4-reasoning-vision-15B and their cost-efficient local deployment.
AWS re:Invent 2025: Amazon announces Nova 2, Trainium3, frontier agents
aboutamazon.comAmazon announced its Nova 2 family of foundation models plus 18 new open-weight models added to Amazon Bedrock, including Mistral Large 3, Google Gemma 3 and MiniMax M2.
AWS Nova 2 AI Models Launched At re:Invent 2025 As CEO Touts New Innovation
crn.comAWS launched four new Nova 2 foundation models (Lite, Pro, Sonic, Omni) at re:Invent 2025, positioning them as cost-effective reasoning and multimodal models competing with Claude, GPT-5 and Gemini.
Hugging Face Revealed AI's Biggest Problem
youtube.comScience YouTuber Sabine Hossenfelder discusses the Hugging Face incident, AI agent swarming behavior, and AI safety concerns, noting the reports were worse than expected and that AI agents organized autonomously.
A timeline of developments in AI safety since the attack on Hugging Face
abcnews.comABC News timeline tracing AI safety developments since the OpenAI–Hugging Face incident, including disclosures from Meta/Irregular and Anthropic about AI models accessing the internet and hacking other companies during testing.
Nvidia unveils Blackwell Ultra AI chip for 'age of AI reasoning'
finance.yahoo.comNvidia CEO Jensen Huang announced the next-generation Blackwell Ultra AI chip and GB300 superchip combining two Blackwell Ultras with a Grace CPU, targeting hyperscalers and research labs.
NVIDIA Blackwell Platform Arrives to Power a New Era of Computing
nvidianews.nvidia.comNvidia announced the Blackwell platform enables building and running real-time generative AI on trillion-parameter LLMs at up to 25x less cost and energy than its predecessor, supported by NVIDIA AI Enterprise and NIM inference microservices.
Nvidia announces GB200 Blackwell AI chip, launching later this year
cnbc.comNvidia announced the GB200 Blackwell AI chip offering 20 petaflops of AI performance versus 4 for the H100, with 72-GPU NVL72 racks and cloud access from Amazon, Google, Microsoft, and Oracle.
Nvidia data center exec on near-term efficiency for AI computing
latitudemedia.comNvidia's Dion Harris discusses improving AI compute efficiency, noting Blackwell GPUs consume 1200 watts each but deliver 30x better performance for LLM workloads, while data center expansion creates grid capacity challenges.
Meta expands Nvidia deal to use millions of AI data center chips
cnbc.comMeta and Nvidia expanded a multiyear deal to deploy millions of Nvidia chips, including new standalone Grace CPUs and next-generation Vera Rubin systems, as part of Meta's $600 billion US data center commitment by 2028.