Other labText & chat
UnconfirmedNotablePareto 26.10 Preview
A preview AI model from an independent developer, able to read pictures and text, for developers building coding tools and automated assistants.
News · earlier edition
Anthropic's IPO demand, a reported push to oust Nvidia's CEO, Google's new Gemini Decision 1 API and OpenAI's $10B quarter led the last 24 hours in AI news.
118 stories ranked · updated Nov 11, 23:02 UTC
Bloomberg sources say Anthropic's IPO drew multiple times the orders for the shares on offer, with a Dario Amodei and Jamie Dimon livestream planned for Friday. The FT reports bankers are weighing a higher IPO price that could lift the valuation target to $1.5T or more, and Australian retail investors can now buy Class A shares.
The WSJ reports Nvidia's board may meet as soon as Wednesday to decide Huang's future after an internal investigation into an affair with an alleged Chinese spy, and the NYT reports a separate relationship with a company executive. Nvidia shares fell 5.47% on Tuesday after a 4%+ drop on Monday.
The API: How developers plug a model into their own apps and pay per use. A model that's "in the API" can be used by programmers, even if it isn't in a chat app yet. returns predefined answers with confidence scores and can score up to 1,000 closed-ended questions in one pass across text, images, audio and video. Google says results come back in 150 to 250 milliseconds.
OpenAI's quarterly revenue grew about 50% to $10B, driven by enterprise sales. Reports put adjusted operating losses at $10.9B, while leaked documents show its adjusted operating margin improved from -150% to -109% between Q2 and Q3.
Anthropic claims its Mythos model moved Claude to AMD chips in about three days with almost no human intervention, which it presents as eroding Nvidia's CUDA lead. The claim comes from Anthropic and has not been independently verified.
AlphaGenome 2 predicts how DNA sequences and their variations affect specific human cell types and disease. It was trained on over 100M cell profiles across 68 tissues and is released for non-commercial research.
Microsoft AI posted pages for MAI-Cyber-1-Flash, MAI-Image-2-6, MAI-Voice-2 and a custom-model tuning offering. Launch details are limited, and the posts' text was not available.
United says it will prevent personal AI assistants from booking tickets on its site, citing privacy risks and possible FAA and DOT rule violations. The airline says it has not yet seen many people try to use AI Agent: An AI that doesn't just answer but takes steps on its own to finish a task: searching the web, running code or using apps. for bookings.
Researchers found more than 35,000 apps built with Grok Build were open to anyone with a link by default, exposing patient data, plaintext passwords and some SpaceXSI documents. The cause was a default setting that made "anyone with the link" access the preselected option.
German defense firm Helsing launched Clarity, which uses Agent: An AI that doesn't just answer but takes steps on its own to finish a task: searching the web, running code or using apps. to turn sensor data into target recommendations for weapons systems under human oversight. The company says every decision will carry a human name.
Releases
Other labText & chat
UnconfirmedNotableA preview AI model from an independent developer, able to read pictures and text, for developers building coding tools and automated assistants.
Meta's researchers released StereoSplat, a free-to-download tool that turns two side-by-side camera photos into 3D scenes, for non-commercial research use only.
GoogleText & chat
NotableGoogle's new model that quickly sorts and judges content such as text, images, audio, video and PDFs, for developers building apps.
GoogleSpecialist
NotableGoogle DeepMind's new science tool predicts how DNA changes affect different human cells and diseases, aimed at researchers studying health.
NVIDIA's slimmed-down copy of Moonshot AI's Kimi K3.1 Pro, free to download and built to run on NVIDIA's newest graphics chips for developers building chatbots and AI agents.
Other labText & chat
UnconfirmedMinorOpenA small model from the Nanbeige group may be coming to a popular developer toolkit, possibly free to download and light enough for a laptop.
A Meta research tool that works out how far away objects are from pairs of camera images. It's for non-commercial research and needs access requested first.
Other labText & chat
UnconfirmedNotableOpenA mid-sized text model from Zyphra that may be free to download and light enough for a good laptop, spotted before launch and likely aimed at developers.
Ant Group's small research model, built on Alibaba's Qwen, that judges questions about text and up to four pictures and gives how likely each answer is. Free to download.
MicrosoftVoice & audio
NotableMicrosoft's new voice model turns written text into speech in a custom voice, for people building voice products.
Other labText & chat
UnconfirmedNotableMade by KT, a South Korean telecom company, this text-only reasoning model is available to developers through OpenRouter, a service that offers access to many AI models.
Turn on the daily digest for any channel: email, Telegram, Discord, push or ntfy.
Archive