Releases

Mistral AI released a command-line interface for managing Spaces and developer workflows.

AWS made automated response-quality, safety, task-completion and tool-use evaluation generally available.

OpenAI announced the close of $122 billion in capital at an $852 billion post-money valuation.

Mistral AI secured approximately $830 million in debt financing to buy NVIDIA chips and develop a Paris-area AI data centre.

CEO David Holz said revenue had significantly surpassed $200 million in 2023 and had increased since then.

Meta released an efficiency update adding object multiplexing and up to 32-frame-per-second multi-object video tracking on one H100.

Suno launched preference learning for genres, moods, and musical references.

Suno launched personalized models trained on a creator's own catalog.

Suno launched verified private voice personalization.

Suno released v5.5 with more expressive generation and new personalization capabilities.

Shield AI announced an agreement to acquire simulation-software company Aechelon Technology.

Shield AI announced $1.5 billion in Series G equity and $500 million in preferred financing at a $12.7 billion post-money valuation.

Cohere entered speech recognition with the open Transcribe model family.

Cohere introduced an open-source speech-recognition model for accurate multilingual transcription.

Perplexity and Samsung expanded AI browsing and assistant experiences across Samsung devices.

Meta released a tri-modal foundation model that predicts high-resolution human brain activity from video, audio and language.

Sierra introduced an agent that creates, executes and continuously improves other customer-experience agents from described outcomes.

Harvey raised $200 million at an $11 billion valuation, co-led by GIC and Sequoia.

Harvey introduced enterprise governance controls for secure AI adoption and administration.

Oracle added private agent creation, lakehouse vectors and secure agent connectivity across its AI Database portfolio.

Oracle introduced an MCP server for secure agent access to Autonomous AI Database.

Oracle added native vector indexing and search over Apache Iceberg tables.

Oracle introduced a no-code system for building private agents in cloud or on-premises environments.

Oracle introduced a developer-focused vector database built on Autonomous AI Database.

Databricks announced Lakewatch, an agentic SIEM for detection, investigation and response on a security lakehouse.

Scale Labs updated MultiChallenge with a stronger judge model, refined tasks, and refreshed results to improve agreement with expert ratings.

Mistral AI released Voxtral TTS, a 4B multilingual text-to-speech model.

AWS released developer tooling for building customized Nova-based frontier models.

AMD Ventures announced an investment and collaboration with World Labs around spatial intelligence, world models and AMD compute infrastructure.

Skild announced partnerships to deploy the Skild Brain across ABB and Universal Robots systems and on dual-arm robots assembling NVIDIA Blackwell systems with Foxconn.

Perplexity launched a health information and research experience using premium medical sources.

Perplexity formed a Health Advisory Board to guide safety, quality and product development for health information.

Harvey launched an integration for analyzing Box-hosted legal documents.

Perplexity launched Comet for iOS, bringing its AI-native browser to all major consumer platforms.

Midjourney released an early-access alpha of its V8 image model.

Perplexity launched Comet Enterprise with managed deployment, policy controls, audit logs and security integrations.

Mistral AI launched Forge, an enterprise system for creating organization-specific frontier models.

OpenAI released GPT-5.4 mini and GPT-5.4 nano as smaller, lower-latency variants.

Crusoe expanded its NVIDIA collaboration across AI-factory design, Vera Rubin systems, model development and inference infrastructure.

Crusoe launched Serverless Fine-Tuning in private preview with OpenAI-compatible APIs and managed model-lifecycle tooling.

Crusoe made Command Center generally available for orchestration, observability and support across AI clusters.

CoreWeave announced platform enhancements across infrastructure, orchestration, inference, developer tools, and operational services for production-scale AI.

Cohere and NVIDIA expanded technical work on efficient, privately deployable sovereign AI.

Scale integrated its Physical AI Data Engine with Universal Robots’ AI Trainer to collect production-grade visual and force-feedback data on industrial robots.

Perplexity joined NVIDIA's Nemotron Coalition to collaborate on open frontier models.

Mistral AI released Leanstral, an open model for formal theorem proving and Lean code.

Mistral AI released Mistral Small 4, a next-generation compact open model for multimodal and agentic workloads.

Mistral AI joined NVIDIA's Nemotron Coalition to advance open frontier models and shared AI infrastructure.

SpaceXAI made its natural-sounding text-to-speech API generally available.

Microsoft and NVIDIA expand integration across Foundry agents, Azure AI infrastructure, NVIDIA accelerators and physical-AI workflows.

NVIDIA formed a coalition of AI labs to develop an open frontier model that will underpin the upcoming Nemotron 4 family.

NVIDIA introduced Nemotron 3 Omni and VoiceChat, GR00T N1.7, Alpamayo 1.5 and Proteina-Complexa.

NVIDIA announced NemoClaw, combining Nemotron models and OpenShell in a governed stack for OpenClaw agents.

NVIDIA launched open software, models, blueprints and agent skills for enterprise agent development.

NVIDIA released its Vera Rubin DSX AI factory reference design and made the Omniverse DSX Blueprint generally available.

NVIDIA announced seven Vera Rubin platform chips in full production for agentic AI factories.

Crusoe announced Edge Zones, extending Crusoe Cloud onto Spark systems for low-latency and sovereign AI deployments.

Palantir and NVIDIA announced an integrated sovereign AI operating system combining NVIDIA accelerated computing and Palantir software.

Perplexity introduced Premium Sources, adding licensed high-value information sources to its research products.

Perplexity launched an enterprise edition of Computer with organizational security, governance and administration.

Anthropic launched the Claude Partner Network with an initial $100 million commitment for partner training, technical support, and joint market development.

Replit reported more than 50 million users building on the platform.

Replit released Agent 4 as its fastest and most versatile software-building agent.

Replit raised $400 million at a $9 billion valuation.

Databricks introduced Genie Code, an autonomous AI partner for data work, pipeline maintenance and model operations.

Perplexity launched Sandbox API for isolated code execution by AI agents.

Perplexity upgraded Search API with improved extraction and dynamic evaluation benchmarks.

Perplexity launched Agent API as a managed runtime for agentic workflows.

Meta published an infrastructure update describing four MTIA chips developed across two years for large-scale AI inference.

NVIDIA released the open 120-billion-parameter Nemotron 3 Super model for complex agentic AI systems.

Legora reported serving 800 law firms and in-house legal teams in more than 50 markets, supporting tens of thousands of lawyers daily.

Legora reported growing from 40 to 400 team members over the preceding year.

Legora raised $550 million in a Series D led by Accel to accelerate U.S. expansion.

Thinking Machines was reported to have grown from about 30 employees to roughly 120 over its first year.

Thinking Machines and NVIDIA announced a multi-year partnership to deploy at least one gigawatt of Vera Rubin systems, alongside an NVIDIA investment.

CoreWeave introduced more flexible cloud-capacity plans designed to match customer workload duration and deployment needs.

Hugging Face launched mutable, Xet-backed S3-like Storage Buckets for checkpoints, logs, traces, and intermediate AI data.

SpaceXAI released Grok 4.20 and a multi-agent variant through its platform.

Meta and World Resources Institute released a DINOv3-based global forest-canopy model with improved accuracy and detail.

Harvey evolved Workflow Builder into Agent Builder for reusable custom legal agents.

Harvey reported hundreds of thousands of agentic Deep Analysis reports and expanded its use in custom agent workflows.

Databricks redesigned AI Search with decoupled storage and compute to support billions of vectors at lower cost.

Scale Labs introduced repository-grounded rubrics for evaluating and reranking candidate software patches without executing tests.

Scale Labs introduced VeRO, a reproducible harness using versioned agent snapshots, controlled budgets, structured traces, and reference procedures.

Scale expanded SEAL into Scale Labs, a broader research organization focused on evaluation, agents, multimodal systems, post-training, deployments, and oversight.

OpenAI released GPT-5.4 and GPT-5.4 Pro for professional work, coding, computer use, and long-context tasks.

Harvey announced an embedded legal-intelligence agent for Microsoft 365 Copilot.

Modal launched Directory Snapshots for capturing and restoring sandbox filesystem state.

Modal added integrated changelog visibility and workflow improvements to its CLI.

CoreWeave announced an agreement to provide AI inference infrastructure for Perplexity.

Cohere joined Aston Martin Aramco as official generative-AI partner to support engineering and business innovation.

Scale launched SWE Atlas, initially releasing Codebase QnA to evaluate how coding agents investigate and reason about real software systems.

Microsoft Research releases Phi-4 Reasoning Vision, a 15-billion-parameter open-weight multimodal reasoning model.

Perplexity expanded Computer as the unifying interface for search, research, creation and actions.

AWS made centralized policy controls for agent-tool interactions generally available.

Google launches Gemini 3.1 Flash-Lite for low-cost, high-throughput multimodal workloads.

Moonshot published its distributed model-checkpointing engine.

Moonshot published optimized kernels for Kimi Delta Attention.

Moonshot released its expert-parallel communication library for large MoE training and inference.

Moonshot published Mixture-of-Block Attention code and research for efficient long-context modeling.