Releases

Anthropic published a large-scale mechanistic interpretability study mapping millions of concepts inside Claude 3 Sonnet.

Hugging Face launched LeRobot to bring open models, datasets, policies, and reproducible tooling to real-world robotics.

Microsoft Research introduces Aurora, a foundation model for weather and atmospheric forecasting across multiple environmental domains.

Microsoft introduces Phi Silica, an on-device small language model optimized for Windows neural processing units.

Microsoft announces Recall, an on-device searchable visual history for Copilot+ PCs with user privacy controls.

Microsoft introduces Copilot+ PCs with neural processing units and a new generation of on-device AI experiences.

DeepSeek upgraded deepseek-chat with improved instruction following, JSON output and RAG performance.

CoreWeave signed a $7.5 billion debt facility led by Blackstone and Magnetar to expand its contracted high-performance compute fleet.

Meta introduced an early-fusion multimodal foundation model capable of generating and reasoning over interleaved text and images.

Baidu introduced its Autonomous Driving Foundation Model for Level 4 driving systems.

Grok-1.5 rolled out to Grok users on X.

Anthropic appointed Instagram co-founder Mike Krieger as Chief Product Officer.

Google launches AI Overviews in the United States, integrating generative summaries and supporting web links into Search.

Google introduces Imagen 3 with improved photorealism, prompt understanding and text rendering.

Google announces Trillium, its sixth-generation TPU, with major gains in compute performance and energy efficiency.

Google DeepMind demonstrates Project Astra, a real-time multimodal assistant prototype that perceives and reasons across continuous context.

Google DeepMind introduces Veo, a high-definition generative video model with cinematic controls.

Google releases PaliGemma, an open vision-language model for image-text tasks.

Google expands Gemini 1.5 Pro to a two-million-token context window in private preview.

Google introduces Gemini 1.5 Flash, a fast and efficient multimodal model with a one-million-token context window.

Anthropic expanded Claude.ai and its mobile app to users and businesses across Europe.

OpenAI released GPT-4o, an omni model designed for real-time reasoning across audio, vision, and text.

CoreWeave established its European headquarters in London and announced a $1.25 billion regional infrastructure investment plan.

Google DeepMind launches AlphaFold Server for free research access to AlphaFold 3 predictions.

Google DeepMind and Isomorphic Labs introduce AlphaFold 3 for predicting structures and interactions across biomolecules.

Red Hat expanded OpenShift AI as the hybrid-cloud operating layer for model development and serving.

Red Hat and IBM introduced InstructLab for community-driven model skills and alignment.

Red Hat introduced RHEL AI as a foundation-model platform for developing, testing and running open models.

DeepSeek opened an OpenAI-compatible API for its chat and coding models.

Oracle announced a generative coding assistant optimized for Java, SQL and OCI development.

DeepSeek published DeepSeek-V2 base and chat weights with 128K context.

DeepSeek introduced its efficient second-generation mixture-of-experts architecture.

Oracle expanded natural-language-to-SQL and retrieval capabilities for Autonomous Database.

Oracle repositioned its flagship database around integrated AI for enterprise data.

Oracle released native vector generation, indexing and semantic similarity search alongside SQL.

Oracle released Database 23ai with integrated vector search and AI-focused developer capabilities.

Harvey launched Vault for secure analysis of large legal document collections.

CoreWeave raised $1.1 billion in Series C funding led by Coatue to expand its AI cloud across products and geographies.

Anthropic launched its first native Claude application for iOS.

Anthropic launched the Claude Team plan for collaborative organizational use.

Amazon released its second-generation multilingual text embedding model through Bedrock.

AWS folded Amazon CodeWhisperer into Amazon Q Developer while preserving its coding-assistance capabilities.

AWS made its software-development assistant generally available and incorporated CodeWhisperer capabilities.

AWS made its enterprise generative-AI assistant generally available.

GitHub opens a technical preview of Copilot Workspace, an agentic environment for moving from issues to plans and implementations.

IBM agreed to acquire HashiCorp to expand hybrid-cloud automation for infrastructure supporting generative AI.

NVIDIA entered a definitive agreement to acquire AI workload orchestration provider Run:ai.

Perplexity raised $62.7 million in a Series B extension led by Daniel Gross, taking its valuation above $1 billion.

Perplexity launched Enterprise Pro, its first business subscription with organizational search, security and data controls.

AWS made configurable Bedrock safety controls generally available.

Microsoft releases Phi-3 Mini, a 3.8-billion-parameter model designed for capable language and reasoning on constrained devices.

Meta and World Resources Institute released global high-resolution canopy-height maps built with DINOv2.

Meta released the Llama 3-based update to its open safeguard model.

Meta released the 8B Llama 3 pretrained and instruction-tuned models.

Meta released the 70B Llama 3 pretrained and instruction-tuned models.

Hugging Face released IDEFICS2, an 8B visual-language model with stronger OCR, document, and visual reasoning capabilities.

Mistral AI released Mixtral 8x22B, a larger open sparse mixture-of-experts model.

Microsoft invests $1.5 billion in G42 and expands a strategic partnership covering Azure, responsible AI, infrastructure and emerging-market deployments.

xAI previewed its first multimodal model, adding understanding of documents, diagrams, charts, screenshots and photographs.

Meta introduced a second-generation MTIA accelerator with increased performance and memory capacity.

Google releases CodeGemma variants for code completion, code generation and instruction following.

Databricks optimized DBRX inference and made the model available through Mosaic AI Model Serving.

Oracle and Palantir partnered to make Palantir platforms available across Oracle Cloud Infrastructure and combine cloud, data, and AI capabilities.

Command R+ became available first through Microsoft Azure alongside Cohere's hosted API.

Cohere introduced Command R+, its higher-capacity enterprise model for advanced RAG, multilingual generation and tools.

Replit introduced a native 7B model specialized for fixing language-server diagnostic errors.

Anthropic published research showing how long-context models can be jailbroken using many demonstrations and proposed mitigations.

Sentient's ViDAR visual detection and ranging capability joined Shield AI's product portfolio.

Shield AI acquired Sentient Vision Systems and its wide-area motion-imaging portfolio.

Security Copilot becomes generally available with integrations across Microsoft's security portfolio.

xAI announced Grok-1.5 with improved reasoning and a 128,000-token context window.

Databricks released DBRX, a 132-billion-total-parameter mixture-of-experts model with 36 billion active parameters under an open license.

Amazon invested an additional $2.75 billion in Anthropic, completing the $4 billion commitment announced in 2023.

Amazon invested an additional $2.75 billion, completing its initial $4 billion Anthropic investment commitment.

Cohere Labs made Command R weights available for research and evaluation.

Cohere released Command R, a 128K-context model optimized for production RAG, citations and tool use.

Suno released v3 with improved full-song generation, vocals, and instrumentation.

Midjourney released new Style Reference algorithms and versioning.

Anthropic, AWS, and Accenture partnered to help enterprises, including regulated industries, deploy Claude-based solutions.

Harvey launched its professional-services AI platform on Microsoft Azure.

Microsoft enters an AI licensing arrangement with Inflection and hires Mustafa Suleyman, Karén Simonyan and members of the Inflection team to build Microsoft AI.

Microsoft forms the Microsoft AI organization focused on Copilot and consumer AI, appointing Mustafa Suleyman as EVP and CEO and Karén Simonyan as Chief Scientist.

Anthropic made Claude 3 models available to Google Cloud customers through Vertex AI.

Kimi opened a beta supporting up to two million Chinese characters of context.

Microsoft and NVIDIA expand integrations across Azure AI, NVIDIA NIM, Omniverse and enterprise generative-AI infrastructure.

NVIDIA and AWS announced broader Blackwell infrastructure and generative AI software integrations.

NVIDIA and Microsoft expanded integrations spanning Azure AI, NIM, Omniverse and enterprise generative AI.

NVIDIA introduced the Blackwell GPU architecture, B100, B200 and GB200 Grace Blackwell Superchip for generative AI.

NVIDIA launched NIM, containerized optimized inference microservices for generative AI deployment.

NVIDIA announced DGX GB200 and a Blackwell-powered DGX SuperPOD reference architecture.

NVIDIA launched Earth-2 cloud APIs and AI models for high-resolution weather and climate simulation.

NVIDIA announced a general-purpose foundation model initiative for humanoid robots.

xAI released the 314B-parameter Grok-1 base weights and Mixture-of-Experts architecture under Apache 2.0.

Google DeepMind introduces SIMA, an instructable agent that follows natural-language directions across multiple 3D environments.

Anthropic released Claude 3 Haiku as the fastest and least expensive model in the Claude 3 family.

Midjourney introduced Character Reference for preserving character appearance across generations.

Oracle introduced a cloud platform combining clinical, claims and operational data with analytics and AI.

Amazon researchers released pretrained Chronos-T5 time-series foundation models and code.

DeepSeek introduced an open vision-language model family for real-world multimodal understanding.

DeepSeek released 1.3B and 7B base and chat checkpoints.