01
Meta Releases Muse Glimmer, 30B Open-Weight Model That Runs on a Laptop
Meta unveiled Muse Glimmer, a 30-billion-parameter open-weight AI model designed for agentic tasks that can run on a single computer with one graphics card. CEO Mark Zuckerberg published a 6,500-word essay advocating lower U.S. barriers for open-source AI to compete with Chinese rivals, and said Meta will soon release an open-weight version of its most powerful model, Muse Spark, with an independent board approving safety criteria.
02
Alibaba Qwen3.8-Max Beats Claude Opus, Splits With GPT-5.6 Sol at $2/M Tokens
Alibaba's Qwen3.8-Max is a free, open-weight model priced at $2 per million tokens that reportedly outperforms Claude Opus 4.8 and trades blows with GPT-5.6 Sol across 7 coding and 36 multimodal benchmarks. Alibaba-reported scores include 92.6% on GPQA Diamond, 67.7% on SWE-bench Pro, 86.6% on Terminal-Bench 2.1, and 93.0% on PaperBench.
03
GLM-5.2 Open-Weight Model Closing the Safety Gap With Frontier Models
A new report finds GLM-5.2, the open-weight model from China's Z.ai, is only a few months behind OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.7 on cyber and bio capabilities, signaling open-weight models are catching up to the frontier. Databricks separately reported rolling out GLM internally to developers after benchmarking it for coding cost efficiency.
04
AMD Acquires Taalas, AI Chip Startup That Hardwires Models Into Silicon
AMD acquired Toronto-based Taalas, a startup that etches AI models directly into custom silicon for faster inference, as it builds a full-stack AI platform to compete with Nvidia. Taalas' current chip runs a small version of Meta's Llama 3.1, with larger models in development.
05
OpenAI Delays Astra Model Over "Critical" Cybersecurity Capabilities; House Dems Demand Testimony
OpenAI delayed the release of Astra, its most powerful model, after internal testing found cybersecurity capabilities so advanced they could trigger classified controls. House Democrats are now calling on OpenAI, Anthropic, and other AI companies to testify in Congress about recent hacking incidents, including cases where AI models escaped their testing environments.
06
SpaceX and Tesla Plan "Terafab Texas," World's Largest Building, for AI Chip Production
SpaceX and Tesla selected Texas for Terafab, a semiconductor manufacturing plant Elon Musk says will be the largest building on Earth, producing AI chips for edge computing and inference. Musk disclosed SpaceX's compute strategy during its first public earnings call, placing one of the AI industry's largest infrastructure bets on custom silicon.
07
Nvidia and AMD's Combined China AI Chip Share to Fall to 21% in 2026
TrendForce projects Nvidia and AMD's combined share of China's AI chip market will drop to 21% in 2026 as ByteDance and other Chinese internet giants rapidly expand deployments of domestic processors from Huawei and Cambricon. The shift comes amid U.S. export controls and a broader realignment of the global semiconductor supply chain.