Releases

CoreWeave introduced ARENA, a production-scale lab for testing AI workload performance, economics, and operational readiness before deployment.

Perplexity launched Model Council, allowing users to query multiple frontier models and synthesize their agreements and disagreements.

Anthropic released Claude Opus 4.6 with stronger coding, longer-running agentic work, improved knowledge work, and a one-million-token context window in beta.

OpenAI released GPT-5.3-Codex, an interactive agentic coding model for complex software and professional work.

Mistral AI released two second-generation Voxtral transcription models for production speech-to-text workloads.

SpaceX acquired X.AI Holdings in an all-stock merger. Reporting tied to the transaction valued xAI at about $250 billion and SpaceX at $1 trillion, implying a combined $1.25 trillion value.

Anthropic named the Allen Institute and Howard Hughes Medical Institute as founding life-sciences partners to develop specialized AI agents and accelerate scientific discovery.

Moonshot opened public beta access to Kimi Claw, a persistent cloud agent workspace.

Baidu released an upgraded 0.9B document-parsing vision-language model.

Modal added richer observability and management controls for running Sandboxes.

Modal became available through the AWS and Google Cloud marketplaces.

Cohere launched Model Vault for dedicated, fully managed and isolated model inference.

xAI launched next-generation image and video generation through the Grok Imagine API, alongside a Batch API for high-volume workloads.

Moonshot released Kimi K2.5, adding native multimodal capabilities and stronger agentic coding.

DeepSeek published its second-generation visual causal-flow OCR model.

Agent Bricks Knowledge Assistant became generally available, powered by instructed retrieval and MLflow evaluation.

Mistral AI expanded Vibe into a fuller agentic development environment for software teams.

NVIDIA agreed to test and validate CoreWeave’s platform software and reference architectures to accelerate deployment of AI factories.

Scale researchers released LHAW, a framework for generating underspecified long-horizon tasks and evaluating whether agents clarify ambiguity and recover performance.

Scale reported that 2025 was its strongest financial year, with more than $1 billion in new business, profitable data operations, and accelerating applications and government work.

Google releases FunctionGemma, a compact open model specialized for function calling and on-device agent workflows.

Anthropic expanded Labs as a dedicated team for incubating experimental products at the frontier of Claude's capabilities.

Anthropic released Cowork as a research preview, bringing Claude's agentic capabilities to desktop work across files and applications.

Anthropic introduced HIPAA-ready Claude tools and resources for healthcare providers, payers, health-tech companies, and individuals, alongside expanded life-sciences capabilities.

Perplexity launched a specialized offering for public-safety organizations.

xAI closed an upsized $20 billion Series E, above its $15 billion target; contemporaneous reporting placed the company’s valuation at more than $230 billion.

CoreWeave announced it would extend its cloud platform with NVIDIA’s Rubin architecture for the next generation of AI training and inference.

NVIDIA released Cosmos Transfer 2.5, Predict 2.5 and Reason 2, updated GR00T N1.6, and introduced Jetson T4000.

Baidu announced a proposed spin-off and separate listing of its Kunlunxin AI-chip business.

Moonshot completed a reported $500 million financing at a $4.3 billion valuation.

CoreWeave ended 2025 with 43 active data centers, more than 850 MW of active power, and approximately 3.1 GW of contracted power.

xAI published its frontier-AI risk-management framework covering capability evaluation, safeguards and model development practices.

NVIDIA market capitalisation at end of calendar year 2025, up 38% year-over-year. NVIDIA remained the world's most valuable public company, sustained by record data centre AI GPU demand.

xAI launched secure team and enterprise Grok plans with no training on customer data, team administration, SSO and SCIM options.

xAI launched an expanded production Collections API for retrieval-augmented generation directly within the Grok API.

xAI announced selection to deliver frontier AI capabilities for the U.S. Department of War.

Anthropic and the US Department of Energy announced a multi-year partnership supporting energy, life-sciences, and scientific-productivity work across national laboratories.

OpenAI released GPT-5.2-Codex, an agentic coding model optimized for long-horizon software engineering.

Mistral AI released Mistral OCR 3 with improved document parsing and structured extraction.

xAI launched its Voice Agent API for developers building real-time conversational voice systems.

Google launches Gemini 3 Flash as a fast, production-oriented model in the Gemini 3 family.

OpenAI opened app submissions and launched an app directory in ChatGPT.

Databricks announced it was raising more than $4 billion in Series L equity financing at a $134 billion valuation.

Meta released a unified multimodal model for separating sounds from complex mixtures using natural prompts.

NVIDIA introduced Nemotron 3 Nano, Super and Ultra open models and companion reinforcement-learning and evaluation tools.

Cohere launched Rerank 4, its fourth-generation retrieval model family in Pro and Fast tiers.

xAI and El Salvador announced a nationwide AI education program using Grok in the country’s schools.

OpenAI released GPT-5.2 with improvements for professional knowledge work and long-context tasks.

CoreWeave expanded Mission Control with Telemetry Relay entering GA, GPU Straggler Detection in preview, and the Mission Control Agent in preview.

Mistral AI launched Vibe CLI, an agentic command-line coding environment powered by Devstral.