Google releases the Gemini Decisions API, powered by its Gemini Decision 1 model, which scores up to 1,000 yes/no, choice and rubric questions in one pass over text, images, audio, video and PDFs. Google says it returns confidence scores in 150 to 250 ms and charges $0.10 per 1M input tokens.
NVIDIA introduces GR00T 2, a world-action model that predicts future video and robot actions together, pretrained on 402,730 hours of human egocentric video plus 23,500 hours of robot teleoperation. NVIDIA reports 40% zero-shot success on 50 unseen tasks, rising to 95% after a 30-minute fine-tune with five demonstrations. Code is due in December.
MiniMax releases H3-Image, an open-weight image model that generates a 4096 x 4096 image in under 10 seconds and handles up to 16 reference images. Weights are public. MiniMax says pricing runs 50 to 70% of comparable offerings, with launch week 50% off for 4K requests.
Anthropic adds Claude Sheets, Whiteboards, and Claude Data, letting teams build spreadsheets, boards, and dashboards together inside Claude. Interactive connectors let partners such as Adobe, Atlassian, Salesforce, and NetSuite embed apps directly in Claude. Anthropic says its marketplace lists more than 9,000 connectors and plugins, and Claude Meetings is coming soon.
OpenAI acquires Odessia, an AI travel agent built by Sonder founder Francis Davidson, who says he now works for OpenAI on the future of travel. Terms were not disclosed, and Odessia's tool is already shut down. An OpenAI spokesperson confirmed the hire but announced no travel products.
The /swarm command in Google Antigravity runs named agents that split a task, work in parallel, and coordinate on one shared message board, with a verifier checking diffs and tests. Google limits the preview to AI Pro and AI Ultra plans and says every worker is its own agent, so a swarm burns quota far faster than a single conversation.
SandboxAQ's Switch AI puts team members and their agents into shared rooms where every agent reads the same history and resources. Xinyi Zhao, a principal product manager, says the open-source platform lets an engineer's agent diagnose a VM connectivity failure in seconds and ship a fix within an hour. She says isolated agents drop context at every handoff.
Google Cloud releases Gemini for healthcare in U.S. preview, a product it says coordinates clinical and administrative workflows across EHR, claims and care management systems. CEO Thomas Kurian says connectors give read-only access to Epic and Oracle Health records, and the platform is built to support HIPAA compliance.
Adobe says it is adding new generative editing tools to Photoshop, Premiere and Lightroom, announced at its MAX 2026 conference. Premiere's Generative Media tool lets editors generate video, sound and music on the timeline using Firefly plus models from Google, Kling, Runway and Luma. The company says MCP integrations arrive for Creative Cloud customers in early 2027.
Sixteen governments including France, Canada and Kenya, plus companies such as Anthropic, Microsoft, OpenAI and Mistral AI, signed the Paris Call for AI safety and security on 9 November 2026. Signatories commit to common frontier-model evaluation standards, independent third-party testing, fail-safe shutdown mechanisms and confidential whistleblower channels. The declaration remains open to new endorsements.