AI News, Oct 29, 2026: Google launches Gemini 4 Carbon, its new top AI model
The Brief
Thursday, October 29, 2026
Google's new Gemini 4 Carbon leads the day, alongside Microsoft's cloud earnings, Anthropic's spending tools and new voice AI models.
- 01
NVIDIA released Nemotron 3.5 Transcribe, an open model with 4.3B active parameters that transcribes 40 languages and labels speakers, scoring 4.71% mean word error rate.
Why it matters
Free, open transcription could lower costs for meeting notes, captioning and medical record tools, though it works only on recordings, not live audio.
- 02
Microsoft reported Azure quarterly revenue in dollars for the first time, at $33.2 billion, as total revenue reached a record $92.3 billion.
Why it matters
Cloud services power most AI products. Azure's size shows how much money is moving into the infrastructure behind the AI tools people use.
- 03
Google expanded ads in its AI Search features globally, announced alongside Alphabet's third-quarter revenue of $127.6 billion, up 25% from a year earlier.
Why it matters
People will see more ads inside AI-powered Search answers in more countries, and advertisers will shift budgets toward them.
- 04
Microsoft said it earned $10 billion in the quarter ending September 30 from commercial arrangements with OpenAI, including revenue-sharing payments.
Why it matters
Shows how closely tied Microsoft and OpenAI are, which affects the future of Copilot, ChatGPT and the tools many people use at work.
- 05
Anthropic launched tools that show companies how their spending on its models is used, ahead of a planned IPO.
Why it matters
Businesses can see where their Claude costs go, which makes it easier to control AI bills and judge whether the tools are worth it.
- 06Google launches Gemini 4 Carbon, its new top AI modelGoogle DeepMind
Google DeepMind released Gemini 4 Carbon, which scores 59 on the Artificial Analysis Intelligence Index, priced at $2/$10 per million tokens, with early access for vetted cyber defenders.
Why it matters
Google's newest model ranks first on a major benchmark and is aimed at coding and security work, with access limited at first to vetted cyber defenders.
- 07
Google updated Chrome's AI-powered autofill to use information from Gmail, such as flight confirmations and order details, to complete forms with the user's permission.
Why it matters
Chrome can fill in forms with flight, order and tracking details from your inbox, saving typing. It only does so with your permission.
- 08
Cybersecurity startup Cathedral used AI agents to access Anthropic's internal systems via third-party software; Anthropic confirmed sales, HR and IT data access but disputes the depth.
Why it matters
Shows that even leading AI companies can be breached, which matters for anyone whose data sits with AI providers.