01
Meta releases Muse Glimmer, a 30B open-weight agentic model under Apache 2.0
Meta released Muse Glimmer, a 30-billion-parameter open-weight model designed to run AI agents locally on a single consumer GPU (24 GB VRAM). Weights are available on Hugging Face under the permissive Apache 2.0 license, and Meta says the model decodes 3.1x faster using DFlash speculation. Mark Zuckerberg published a 6,500-word essay alongside the release arguing that decentralizing AI is safer than concentrating control.
02
Meta's Muse Spark 1.2 scores 54 on Artificial Analysis Intelligence Index, open weights promised soon
Meta's more powerful closed model, Muse Spark 1.2, reached a score of 54 on the Artificial Analysis Intelligence Index (AAII), roughly matching Grok 4.5 and jumping significantly in four months. Zuckerberg said Meta will open the weights for Spark 1.2 in the coming weeks, while the smaller Glimmer is already freely downloadable.
03
Alibaba's open-weight Qwen3.8-Max beats Claude Opus 4.8 on benchmarks at $2 per million tokens
Alibaba released Qwen3.8-Max as a free, open-weight model that scores 56 on the Artificial Analysis Intelligence Index (a 10-point jump over Qwen3.7 Max) and claims to beat Claude Opus 4.8 on key benchmarks while splitting with GPT-5.6 Sol. Priced at $2 per million tokens via API, it undercuts most closed competitors.
04
Open-weight GLM-5.2 nears frontier on cyber and bio capabilities, but safety mitigations lag
A SaferAI evaluation found Z.ai's open-weight GLM-5.2 is only months behind OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.7 on cyber and bio capability benchmarks, but with significantly fewer safety mitigations. The report highlights a growing gap between open-weight capability and open-weight safety as Chinese labs release powerful models quickly.
05
Nvidia lines up $500 billion in AI infrastructure financing with six Wall Street giants
Nvidia signed memorandums of understanding with Blackstone, BlackRock, Goldman Sachs, Apollo Global, KKR, and Brookfield to create financing platforms that could channel more than $500 billion into AI computing infrastructure over time. The deal marks a shift toward treating AI chips as an investable asset class backed by major financial institutions.
06
Microsoft targets 300,000 next-gen Maia AI chips in 2027, in talks with TSMC
Microsoft is in talks with TSMC to secure capacity for more than 300,000 next-generation Maia AI chips for delivery in 2027, aiming to ramp up internal silicon production to cut reliance on Nvidia. The company is reportedly trying to persuade major cloud customers like Anthropic to adopt its custom silicon.
07
White House AI safety framework exempts open-weight models but keeps details secret
The Trump administration's voluntary AI testing framework will exempt lower-cost open-weight models from safety reviews, applying only to "state-of-the-art" closed-source models deemed national security risks. The framework details remain secret and have not been made public, drawing criticism from transparency advocates even as the open-source exemption is welcomed by the open AI community.