Releases

Anthropic released a web search tool for developers building Claude-powered applications.

Modal released the 1.0 generation of its Python client and SDK.

Google introduces the open Agent2Agent protocol with more than 50 partners to enable interoperable agent collaboration.

Google Cloud releases the open-source Agent Development Kit for building and orchestrating multi-agent systems.

xAI made Grok 3 flagship models generally available through its API.

xAI made its image-generation model available through the developer API.

NVIDIA launched Dynamo to orchestrate and scale distributed generative AI inference.

Anthropic released API features designed to reduce token use and improve long-running agent workflows.

Databricks released APIs for embedding Genie conversational analytics into applications, Teams, Slack and SharePoint.

OpenAI released the open-source Agents SDK for orchestrating single- and multi-agent workflows.

OpenAI launched the Responses API as a new foundation for building agentic applications with built-in tools.

Hugging Face launched a unified inference layer that lets developers call models across multiple providers through one API and billing experience.

Perplexity launched Sonar and made Sonar Pro generally available through its API platform for cited real-time answers.

Hugging Face released smolagents, a lightweight open-source framework for tool-using and code-executing AI agents.

Anthropic introduced and open-sourced MCP, its specification, SDKs, local Claude Desktop support, and reference servers.

Mistral AI launched a moderation API for safety classification and content policy enforcement.

Mistral AI launched an asynchronous Batch API for high-volume model inference.

xAI opened public API access to Grok foundation models with OpenAI- and Anthropic-compatible interfaces.

Gradio 5 introduced redesigned application theming, stronger security defaults, and improved AI app development workflows.

Anthropic launched computer use in public beta, allowing developers to direct Claude to operate graphical computer interfaces.

Anthropic launched asynchronous batch processing for large volumes of Claude API requests.

OpenAI launched the Realtime API public beta for low-latency multimodal experiences.

Anthropic introduced Contextual Retrieval, combining contextual embeddings and BM25 to improve retrieval-augmented generation.

La Plateforme added support for Pixtral and expanded developer tooling for multimodal applications.

Accelerate reached its 1.0 release-candidate milestone after becoming a core distributed-compute layer in the Hugging Face ecosystem.

Anthropic launched prompt caching to reduce latency and cost for repeated long-context prompts.

DeepSeek introduced disk-based context caching to reduce repeated-token costs.

The API gained JSON mode, function calling, chat-prefix completion and fill-in-the-middle completion.

Mistral AI introduced self-service fine-tuning through La Plateforme and released customization tooling.

Anthropic made tool use generally available across the Claude API, Amazon Bedrock, and Google Vertex AI.

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

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

Cohere added function calling and tool-use capabilities for building grounded enterprise agents.

Anthropic expanded commercial legal protections and improved API access and enterprise terms.

Google releases the Gemini API through Google AI Studio and Vertex AI, giving developers access to Gemini Pro.

Baidu fully opened ERNIE APIs to enterprise customers through AI Cloud.

xAI began early API access for Grok developers ahead of the later public beta.

OpenAI launched the Assistants API beta with Code Interpreter, Retrieval, and function calling.

Moonshot introduced developer API access to its language models.

Hugging Face released a production-grade server for serving text embedding and reranking models.

Meta introduced ExecuTorch, an open-source framework for efficient on-device PyTorch inference.

NVIDIA publicly released its open-source TensorRT-LLM library for optimized large-model inference.

HTTP endpoints backed by autoscaling Modal functions became generally available.

Modal's Python-first SDK and developer workflow reached general availability.

Perplexity launched pplx-api with hosted access to its PPLX chat models.

Microsoft Research releases AutoGen, a framework for building applications with conversing and collaborating AI agents.

Databricks introduced pay-per-token APIs and provisioned throughput for serving curated open foundation models.

Hugging Face released Candle, a minimalist Rust framework for fast and portable machine learning.

Meta released AudioCraft with MusicGen, AudioGen and EnCodec components for generative audio research.

CoreWeave released Tensorizer, an open-source tensor serialization tool that streams models from S3, HTTP, and file storage to reduce PyTorch model startup time.