Releases

Mistral AI released Mistral Small 3.2 with improved instruction following, function calling and reduced repetition.

Meta and EssilorLuxottica introduced performance-focused Oakley Meta HSTN AI glasses.

Databricks introduced managed Model Context Protocol servers for Unity Catalog functions, Genie spaces and AI Search indexes.

Google introduces Gemini 2.5 Flash-Lite as a high-throughput, low-cost model in preview.

Google makes Gemini 2.5 Pro and Gemini 2.5 Flash generally available for developers and enterprises.

Moonshot released a 72B software-engineering model optimized using reinforcement learning.

Databricks introduced a simplified business-user front door to Genie spaces, dashboards and apps.

AI/BI Genie became generally available across AWS, Azure and Google Cloud, alongside a new knowledge store and planned Deep Research mode.

Meta made a $14.3 billion strategic investment for a 49% non-voting minority stake in Scale AI, valuing the independent company at over $29 billion and expanding the commercial partnership.

Founder and CEO Alexandr Wang left day-to-day management to join Meta’s AI efforts while remaining a Scale director; Chief Strategy Officer Jason Droege became interim CEO.

Perplexity closed a $500 million financing led by Accel at a $14 billion valuation.

Meta made a $14.3 billion non-voting minority investment in Scale AI and expanded the companies' commercial relationship; Scale was valued above $29 billion.

Databricks and Microsoft expanded integration across Azure Databricks, Microsoft 365 and governed enterprise agents.

Databricks released MLflow 3 with unified tracking, evaluation, observability and governance for models and agents.

Databricks introduced Agent Bricks, which generates task-specific data and benchmarks to optimize agents for enterprise use cases.

Databricks introduced Lakebase, a serverless Postgres category designed for operational data and AI agents.

Hugging Face and NVIDIA introduced a managed service for research organizations to access large GPU clusters for foundation-model training.

Perplexity partnered with European model providers to bring additional European AI models to global users.

Mistral AI announced Mistral Compute, a sovereign European AI cloud developed with NVIDIA infrastructure and European partners.

Meta released V-JEPA 2 for video understanding, physical-world prediction and zero-shot robot planning.

NVIDIA expanded DGX Cloud Lepton with European cloud providers, AWS and Azure participation, and Hugging Face training integration.

Perplexity published work on accelerating Sonar inference through speculative decoding.

Mistral AI released Magistral Small and Magistral Medium, its first reasoning-focused model family.

OpenAI released o3-pro for more reliable reasoning by using additional compute.

Cohere announced work with the UK government to support secure AI adoption and public-sector productivity.

Cohere and the Government of Canada agreed to explore secure generative AI for public-sector operations.

xAI released a Management API for programmatic API-key administration.

Anthropic expanded access and deployment support for approved US national-security customers.

Perplexity expanded Finance with richer market data and research tools for investors.

The companies submitted the largest NVIDIA Blackwell MLPerf Training v5.0 cluster and reported record Llama 3.1 405B training performance.

Mistral AI launched Mistral Code, an enterprise coding-assistant platform with IDE integration and deployment controls.

Former Meta FAIR head Joelle Pineau joined Cohere as Chief AI Officer to lead research and model development.

Hugging Face released SmolVLA, a 450M vision-language-action model trained for affordable, open robotic manipulation.

Modal introduced interactive notebooks backed by elastic cloud compute.

Modal added snapshots to accelerate reproducible sandbox and agent environments.

Moonshot published a seven-billion-parameter formal theorem-proving model.

Scale formalized a data-engine offering for robotics and embodied AI, covering multimodal collection, annotation, and real-world model improvement.

CoreWeave made eight-GPU NVIDIA HGX B200 Blackwell instances generally available for enterprise AI training and inference.

Perplexity launched Labs for creating reports, spreadsheets, dashboards and lightweight applications from research.

DeepSeek upgraded R1 with stronger reasoning, reduced hallucination and JSON/function-calling support.

Mistral AI released Codestral Embed for semantic retrieval and search across software repositories.

NVIDIA reported Q1 fiscal 2026 revenue of $44.1 billion, up 69% year-over-year and 12% sequentially. Data Center revenue was $39.1 billion (+73% YoY). Net income was $18.8 billion (GAAP). Results included a $4.5 billion charge related to H20 export control restrictions.

Anthropic appointed Netflix co-founder Reed Hastings to its board of directors.

Mistral AI launched the Agents API with persistent agents, tools, orchestration, conversations and built-in connectors.

Anthropic activated ASL-3 deployment and security protections alongside the launch of Claude Opus 4.

Anthropic moved Claude Code from research preview to general availability and expanded its integrations.

Anthropic released Claude Opus 4 and Claude Sonnet 4 for coding, advanced reasoning, and agent workflows.

Perplexity expanded free and paid plan capabilities, including greater access to advanced search, files and models.

Mistral AI and All Hands AI released Devstral Small 1, an open model designed for agentic software engineering.

OpenAI added remote MCP server support, image generation, Code Interpreter, and background mode to the Responses API.