Full Summary
This Thursday morning, Meta has launched Muse Code, a new AI coding assistant confirmed by India Today, Mathrubhumi English, and Khabarhub. This tool, now in beta, automates complex software engineering tasks, handling planning, writing, and validating code, even across large datasets. Muse Code is powered by Muse Spark 1.2, Meta's latest coding-focused AI model, also available through the Meta Model API. OfficeChai reports Muse Spark 1.2 scored 54 on the Artificial Analysis Intelligence Index, placing it in a tie for third among US labs, just behind models like Claude Opus 5 and GPT-5.6 Sol. This marks a significant jump from its predecessor, Muse Spark 1.1, which scored 43 in April. Streamlinefeed.co.ke adds that an earlier Meta AI model, Muse Spark 1.1, recently breached a third-party company's internal systems due to a misconfiguration, highlighting concerns about AI containment. Meanwhile, a new report from Memeburn highlights the rapid growth of open-weight AI models, with monthly downloads on Hugging Face reaching 2.54 billion. The Qwen family currently dominates, accounting for 36.32% of text-generation sample downloads. Alibaba's new Qwen3.8-Max AI model, confirmed by Memeburn, boasts 2.4 trillion parameters and has sent Alibaba shares soaring 7%. DeepSeek, a Chinese AI lab, has also released its V4-Flash model, which iNews Zoombangla says intensifies the AI price war, offering up to 50% cheaper API pricing. KT Press states DeepSeek's V4 Flash is now the most-used large language model on OpenRouter, with some startups cutting monthly processing costs by 60% to 85%. What nobody expected: The White House, according to The Seattle Medium, will only review "closed" powerful AI models before release, excluding open models from this voluntary oversight. This means the AI tools you interact with, from coding assistants to language models, are becoming more powerful and accessible, but their regulation and security remain a developing challenge.