AI This Week

OpenAI said it will retire several models from ChatGPT next month, including GPT‑4o, the model some paid users favored for its warm conversation style. The company launched GPT‑4o in May 2024 and faced backlash in August after briefly removing access following the release of GPT‑5, then restoring it and promising advance notice before any permanent retirement. OpenAI now says only 0.1% of users choose GPT‑4o daily and most people use GPT‑5.2, citing recent upgrades to personality, customization, and creative ideation. GPT‑4.1, GPT‑4.1 mini, and o4-mini will also exit ChatGPT. The API stays unchanged.

Mark Zuckerberg has teased a major AI rollout slated for 2026, including “agentic” commerce tools aimed at handling shopping-related tasks for users. The plans point to AI systems that can take actions, not just answer questions, with a focus on workplace and consumer use cases tied to buying and selling. The preview suggests an expanded set of products and integrations designed to move AI from chat into daily transactions and business workflows. Few specifics were released, but the timing and scope signal a broad product push. The announcement adds fresh momentum to the competition among major tech firms racing to ship action-oriented AI agents.

Anthropic has released a new suite of interactive applications that bring its Claude AI model directly into Slack and other workplace tools. The launch integrates the AI heavily into daily business operations. Instead of switching tabs to access the model, employees can now utilize Claude’s capabilities within their existing communication platforms. The system examines context from chat threads to complete tasks and answer queries instantly. This release marks a significant expansion for the company as it targets enterprise users. The suite includes specific tools designed to reduce friction in professional environments. By placing the model inside third-party software, Anthropic aims to capture the workflow market.

The United Arab Emirates has launched a home-grown “sovereign” open AI model in a bid to challenge dominant US and Chinese systems and boost its technological standing. The new model, backed by state resources, is designed to be openly accessible yet tailored to regional needs, reflecting ambitions to diversify the UAE’s economy and reduce dependence on foreign tech giants. Researchers say its performance positions it alongside some of the best existing open AI models, potentially attracting global developers and enterprises. The move underscores growing competition in the AI landscape as nations seek both technological leadership and economic advantage.

Microsoft has revealed its new Maia 200 AI accelerator chip, a piece of hardware the company says is three times more powerful than rivals from Google and Amazon. The chip is an inference powerhouse. It is designed to help AI models apply their knowledge to real-world situations with greater speed and efficiency. Microsoft is already deploying the Maia 200 in its Azure cloud data centers to improve next-generation language models and power services like Copilot. For end users, this could mean faster responses from AI tools. Scientists and developers using Microsoft’s platforms may also see a performance boost for large-scale projects, from advanced weather modeling to biological research.