Anthropic upgraded Claude's voice mode so paid users can switch between Haiku, Sonnet, and Opus mid-conversation instead of every voice query running through Haiku alone; voice mode now defaults to whichever model the user last used in text chat. Free-tier voice remains single-connection and Haiku-only. Separately, fast mode for Claude Opus 4.7 is being retired on July 24 as users are migrated to fast mode on Opus 4.8.
OpenAI made ChatGPT Health — its feature for surfacing and organizing health information — available to all U.S.-based users 18 and older across every plan, on web and iOS. Users can connect supported health records and Apple Health data to see labs, medications, activity, and sleep data in one dashboard alongside ChatGPT's usual chat interface. OpenAI says health-related query volume has grown from 230 million to 300 million per week since testing began.
No significant updates today. Most recent notable news remains Tuesday's (Jul 21) launch of Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and the security-tuned Gemini 3.5 Flash Cyber, plus Google confirming it has begun pretraining Gemini 4. Gemini 3.5 Pro, the flagship, has still not shipped.
Microsoft has begun replacing OpenAI's image-generation models with its own homegrown models across consumer products including PowerPoint and Bing Image Creator. AI chief Mustafa Suleyman said the switch is because the in-house models are "faster, cheaper, higher quality" and drive better user retention — another sign of Microsoft reducing reliance on OpenAI's models for its own product surfaces.
Elon Musk and the xAI account posted a burst of updates on the evening of Jul 23: a new Tavily search plugin for Grok Build (xAI's coding/developer agent platform) that returns LLM-optimized, structured search results instead of raw web pages, plus a fresh tease naming "Grok 4.6" and tying it to a specific benchmark claim — the clearest signal yet of xAI's next release after Grok 4.5, which shipped July 8.
Nvidia signed a multi-year agreement to prepay Amkor Technology $1.5 billion (not an equity stake) to expand advanced semiconductor packaging and test capacity at Amkor's Arizona campus, near TSMC's Arizona fabs. The companies will jointly develop high-density interconnect and heterogeneous-integration packaging for Nvidia's AI and accelerated-computing chips; Amkor shares jumped as much as 17% after-hours on the news.
Nvidia and the Korea Advanced Institute of Science and Technology announced a joint AI research laboratory at KAIST's Kim Jaechul Graduate School of AI, focused on advancing agentic AI for South Korea — part of Nvidia's broader push to embed itself in national AI research programs (also visible in its DGX GB300 commissioning at the Naval Postgraduate School on Jul 22).
The Trump administration selected advanced reactor developers Oklo and X-Energy to join a $200 million Department of Energy program — alongside Microsoft and Nvidia — aimed at compressing the licensing, design, and engineering timelines that stand between new reactor designs and power-hungry AI data centers, with funding routed through DOE national labs and academic partners including UT Austin. Both stocks jumped as much as 10–12% in after-hours trading on the news.
The Department of Energy selected Idaho National Laboratory, the Oak Ridge Reservation (Tennessee), the Paducah Gaseous Diffusion Plant (Kentucky), and the Savannah River Site (South Carolina) to move forward with private-sector partnerships for AI data center and co-located energy generation projects on federal land. It's the next concrete step under the Trump administration's July 23 infrastructure action plan, which aims to streamline permitting and use federal land to speed AI buildout and grid connections, with nuclear power generation a central piece of the sites' pitch.
Lunar Outpost, a startup building robotics for space infrastructure, announced its next lunar rover will run its lidar system on Nvidia Jetson chips — likely making it the first Nvidia GPU to operate on the Moon's surface. It's a small but notable data point in Nvidia's push (alongside its recent "space computing" platform push) to get accelerated-computing hardware qualified for space and lunar environments.