Releases

Meta released SAM 3 for text-, exemplar- and visual-prompted detection, segmentation and tracking in images and video.

NVIDIA reported Q3 fiscal 2026 record revenue of $57.0 billion, up 62% year-over-year and 22% sequentially. Data Center revenue was $51.2 billion (+66% YoY). Net income was $31.9 billion (GAAP). Jensen Huang noted 'Blackwell sales are off the charts, and cloud GPUs are sold out.'

Databricks became an early Microsoft Agent 365 partner, bringing governed Agent Bricks experiences to Teams, Outlook and Excel.

Microsoft and Anthropic expand their partnership, bringing Claude models to Microsoft Foundry and Microsoft 365 Copilot.

Google launches Antigravity, an agent-first development environment built around autonomous planning, coding and verification.

Google launches Gemini 3 Pro for advanced multimodal reasoning, coding and agentic work.

Anthropic and Microsoft expanded their partnership, bringing Claude models to Microsoft Foundry and Microsoft 365 Copilot.

xAI released Grok 4.1 across grok.com, X, iOS and Android with improvements to creative, emotional and collaborative interactions.

Perplexity added stronger visibility, confirmations and user controls for Comet Assistant actions.

Baidu upgraded Qianfan with stronger model services and enterprise agent-development infrastructure.

Baidu introduced a self-evolving machine-learning engineering agent.

OpenAI released GPT-5.1 for developers with improved coding, steerability, and agentic performance.

CoreWeave increased its revolving credit facility from $1.5 billion to $2.5 billion and extended maturity to November 2029.

Baidu released a sparse multimodal reasoning model activating only 3B parameters.

Databricks introduced ai_parse_document for multimodal PDF and document understanding integrated with Agent Bricks, AI Search and AI/BI.

Meta released speech-recognition models covering more than 1,600 languages plus an open corpus for 350 underserved languages.

Moonshot released the reasoning-focused K2 Thinking model for long-horizon agentic tasks.

Perplexity released a redesigned Comet Assistant with improved browsing and task execution.

Google Cloud announces general availability of Ironwood, its seventh-generation TPU for training and inference.

Scale announced expansion across New York, London, Washington, D.C., Doha, and St. Louis to support enterprise, government, and international growth.

AWS and OpenAI announced a seven-year strategic infrastructure agreement for OpenAI workloads using EC2 UltraServers and NVIDIA GPUs.

OpenAI and AWS announced a multi-year strategic partnership under which OpenAI would access AWS compute infrastructure.

Moonshot released its open-source Kimi coding agent and command-line interface.

Perplexity introduced Privacy Snapshot to show users how Comet and its assistant handle data and permissions.

The proposed Core Scientific acquisition was terminated after Core Scientific stockholders did not approve the all-stock merger agreement; the commercial partnership continued.

CoreWeave announced the acquisition of Marimo, creator of an open-source reactive Python notebook, to unify experimentation and production workflows.

Perplexity launched a free AI-powered patent search and analysis experience.

AWS brought online a large Trainium2 compute cluster in Indiana for Anthropic model training.

Amazon announced approximately 14,000 corporate job reductions while restructuring around fewer management layers and increased AI investment.

OpenAI completed a recapitalization establishing OpenAI Group PBC as a public benefit corporation controlled by the nonprofit OpenAI Foundation.

The core Hugging Face Hub Python client reached v1.0 with a redesigned hf CLI, httpx networking, and Xet-based file transfers.

xAI launched Grokipedia, an AI-assisted encyclopedia and knowledge reference product.

Anthropic launched a beta Excel add-in capable of reading, analyzing, modifying, and creating workbooks.

Mistral AI launched AI Studio as an enterprise production platform for building, evaluating, observing, governing and deploying AI applications.

Google Cloud and Anthropic expand their infrastructure partnership, with Anthropic planning access to up to one million TPUs for Claude training and serving.

DeepSeek introduced its open OCR and document-understanding research family based on visual-text compression.

DeepSeek published its optical-compression model for OCR and document understanding.

Amazon released a universal time-series foundation model for univariate, multivariate and covariate-informed forecasting.

Anthropic launched connectors, skills, and domain support for scientific and life-sciences workflows.

Baidu released a compact multilingual vision-language model for document parsing.

Anthropic introduced reusable Skills across Claude apps, Claude Code, and the API.

Oracle announced new embedded agents across finance, HR, supply chain, sales, marketing and service.

Oracle introduced a marketplace for validated partner-built Fusion application agents.

CoreWeave expanded AI Object Storage with a unified dataset across regions, clouds, and on-premises environments and automated pricing designed to lower typical storage costs by more than 75%.

CoreWeave agreed to provide Poolside with a cluster of more than 40,000 NVIDIA GB300 NVL72 GPUs and support its Project Horizon AI campus.

xAI made web search, X search and code execution tools generally available through its API.

Google launches Veo 3.1 with richer audio, stronger narrative control and reference-image guidance.

NVIDIA began shipping DGX Spark personal AI supercomputers to developers.

Anthropic released Claude Haiku 4.5, offering near-frontier coding performance at lower cost and latency.

Oracle expanded its AI infrastructure roadmap to multi-gigawatt clusters and hundreds of thousands of GPUs.

Oracle unveiled its resilient low-latency networking fabric for very large GPU clusters.

Oracle announced a multi-gigawatt AI supercluster architecture targeting up to 800,000 NVIDIA GPUs.

Oracle added an in-database framework for building agents grounded in enterprise data.

Oracle launched an Apache Iceberg-compatible autonomous lakehouse for AI and analytics across clouds.

Oracle released its AI-native flagship database generation with integrated agentic and vector capabilities.

Anthropic and Salesforce expanded their partnership to make Claude a preferred model for regulated Agentforce deployments.

AWS made the managed agent platform generally available across nine regions.

Microsoft AI announces MAI-Image-1, its first image-generation model developed entirely in-house.

AWS launched the conversational agentic teammate at the center of Quick Suite.

AWS launched an agentic workspace combining research, business intelligence and workflow automation.

Google Cloud launches Gemini Enterprise as an enterprise agentic platform connecting models, data, agents and governance.

IBM added lifecycle observability, monitoring and policy controls for enterprise agents.

Red Hat introduced a unified platform generation for model training, inference and agent workloads.

CoreWeave launched a publicly available serverless reinforcement-learning capability jointly built from Weights & Biases and OpenPipe technology.

Cohere formalized a tiered program for cloud, consulting and strategic partners deploying enterprise AI.

Google releases Gemini 2.5 Computer Use in preview for agents that interact with graphical user interfaces.

CoreWeave agreed to acquire Monolith AI and extend its full-stack platform into industrial engineering, simulation, automotive, and manufacturing workflows.

OpenAI moved Codex to general availability and introduced Slack integration, an SDK, and administration features.

OpenAI launched the first interactive apps inside ChatGPT conversations.

OpenAI released the Apps SDK preview for developers to build interactive applications inside ChatGPT.

IBM launched hybrid Mamba/transformer Granite 4.0 models optimized for efficient agentic workloads.

Perplexity broadened public availability of the Comet browser beyond its initial subscriber and waitlist rollout.

IBM introduced commercial Spyre accelerator availability for generative and high-volume inference.

Modal released a beta Go SDK.

Modal released a beta JavaScript and TypeScript SDK.

Modal launched a beta of its second-generation persistent volume architecture.

Moonshot published a 48B Kimi Linear checkpoint.

Moonshot introduced Kimi Linear and Kimi Delta Attention as a hybrid linear-attention approach.

Perplexity announced the initial media partners participating in Comet Plus.

Microsoft introduces the open-source Microsoft Agent Framework, unifying agent-development concepts from Semantic Kernel and AutoGen.

OpenAI released Sora 2, a video and audio generation model with improved physical realism and control.

Modal announced its Series B, reported at $87 million and a $1.1 billion post-money valuation.

DeepSeek released the experimental V3.2 model with sparse attention.

NVIDIA released GR00T N1.6, Cosmos Reason NIM and new Newton physics integration for Isaac Lab.

Anthropic released an SDK exposing the agent-building infrastructure used by Claude Code.

Anthropic launched a refreshed Claude Code terminal interface, checkpoints, background tasks, subagents, and a native VS Code extension.

Anthropic released Claude Sonnet 4.5 with major gains in coding, computer use, agents, reasoning, and mathematics.

Kimi added the OK Computer agent mode for executing computer-oriented tasks.

A third OpenAI agreement added up to $6.5 billion, bringing total announced contract value with OpenAI to approximately $22.4 billion.

Perplexity launched a developer Search API exposing its real-time web index, retrieval and content extraction.

xAI expanded its government offering through GSA OneGov, making Grok reasoning models available to U.S. federal employees at negotiated pricing.

Meta launched Vibes, a feed in the Meta AI app and meta.ai for creating, remixing and sharing AI-generated short videos.

Google DeepMind introduces Gemini Robotics 1.5 and an embodied-reasoning model for more capable robot agents.

Gemini 2.5 Flash-Lite becomes generally available for high-volume, latency-sensitive applications.

A second close added $100 million from BDC and Nexxus Capital Management, bringing Cohere's valuation to approximately $7 billion.

Cohere expanded work with AMD across model deployment, compute infrastructure and internal use of North.

OpenAI, Oracle, and SoftBank announced five new US Stargate data-center sites, moving planned capacity toward nearly 7 gigawatts.

DeepSeek improved language consistency and its code and search agents.

The US General Services Administration added Meta's Llama models to OneGov to support federal-agency AI adoption.

OpenAI and NVIDIA signed a letter of intent to deploy at least 10 gigawatts of NVIDIA systems, with the first Vera Rubin gigawatt targeted for the second half of 2026.