Releases

Modal expanded persistent storage and data-access capabilities for cloud workloads.

Modal improved support for distributed and GPU-intensive training workloads.

Modal expanded GPU availability and lowered prices across supported accelerators.

Mistral AI released Pixtral 12B, its first open-weight multimodal model.

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

Meta and EssilorLuxottica extended their long-term partnership to develop multiple generations of smart-eyewear products.

Microsoft, BlackRock, Global Infrastructure Partners and MGX form an investment partnership to mobilize capital for AI datacenters and supporting energy infrastructure.

Scale and the Center for AI Safety launched an international effort to create a broad expert-written benchmark for frontier AI systems.

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

OpenAI released its first o-series reasoning models, o1-preview and o1-mini.

Oracle introduced a natural-language assistant for exploring and explaining enterprise analytics.

Oracle expanded its analytics platform with AI-generated insights and application-aware data models.

NotebookLM adds Audio Overviews, generating conversational podcast-style discussions grounded in notebook sources.

Oracle integrated HeatWave GenAI and vector processing with multicloud lakehouse data.

Oracle expanded AutoML with larger models, topic modeling, drift detection and improved anomaly detection.

Anthropic added organization workspaces to separate API keys, usage, and access controls.

DeepSeek merged its general and coding capabilities into V2.5 and published weights.

Mosaic AI Agent Framework and Agent Evaluation entered public preview with integrated deployment, evaluation, tracing and governance.

Anthropic launched Claude for Enterprise with expanded context, security, administration, and GitHub integration.

xAI brought its Memphis Colossus training cluster online with 100,000 NVIDIA H100 GPUs after a 122-day build.

Scale and research collaborators introduced EnigmaEval, a multimodal reasoning benchmark built from novel puzzle-competition problems.

The August Command R+ release delivered roughly 50% higher throughput and 25% lower latency with the same hardware footprint.

Cohere updated Command R with improved tool decisions, instruction following, structured-data analysis and robustness.

Amazon hired Covariant's founders and a portion of its team and received a non-exclusive license to its robotic foundation models.

Perplexity partnered with Comcast to provide eligible Xfinity Rewards members a year of Perplexity Pro.

CoreWeave publicly introduced Mission Control as its managed workflow for validating hardware, monitoring cluster health, and maintaining reliable AI infrastructure.

CoreWeave launched NVIDIA H200 Tensor Core GPU instances and said it was the first cloud provider to bring the H200 to market.

IBM unveiled Telum II with upgraded on-chip AI acceleration for next-generation mainframes.

NVIDIA and partners launched reusable reference workflows for enterprise generative AI agents.

Anthropic made Artifacts generally available across Claude plans and added publishing and remixing capabilities.

Hugging Face released IDEFICS3, a more efficient multimodal architecture that became the basis for the SmolVLM family.

Microsoft releases Phi-3.5 Vision for multimodal image and text understanding.

Microsoft releases Phi-3.5 MoE, a mixture-of-experts small language model.

Microsoft releases Phi-3.5 Mini with multilingual support and a long context window.

DeepSeek released an improved formal theorem prover using reinforcement learning and search.

Google begins broad availability of Imagen 3 through ImageFX and later expands access through Gemini and Vertex AI.

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

xAI released Grok-2 and Grok-2 mini in beta on X and announced enterprise API availability.

Hugging Face acquired XetHub to bring content-defined chunking, deduplication, and high-performance storage to large AI repositories.

Mistral AI expanded La Plateforme with model customization and fine-tuning capabilities.

Amazon updated its image model with image conditioning, subject consistency and background-removal capabilities.

Moonshot raised a reported $300 million round that valued the company at $3.3 billion.

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

Google enters a non-exclusive technology licensing agreement with Character.AI and brings co-founders Noam Shazeer and Daniel De Freitas, plus members of their research team, into Google DeepMind.

Scale released MultiChallenge to measure model performance on realistic multi-turn conversation problems involving memory, instruction retention, editing, and self-consistency.

Perplexity partnered with Uber to offer eligible Uber One members a complimentary year of Perplexity Pro.

GitHub launches Models in beta, letting developers compare and experiment with multiple AI models inside GitHub.

Google releases ShieldGemma, open safety classifiers built on Gemma 2 for filtering model inputs and outputs.

Perplexity launched a publisher program with revenue sharing, API access and technology support.

Meta launched AI Studio for creating, sharing and discovering custom AI characters.

Meta released SAM 2 for promptable object segmentation in both images and video.

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

OpenAI launched a limited prototype of SearchGPT with web answers and linked sources.

DeepSeek upgraded the deepseek-coder API to the July Coder V2 checkpoint.

Mistral AI released Mistral Large 2, a 123B flagship model with a 128K context window and improved multilingual and coding performance.

AWS made its generative-AI clinical documentation service generally available.

Meta released a classifier for detecting prompt injection and jailbreak attempts.

Meta released a multilingual Llama 3.1-based safeguard model.

Meta released its 405B-parameter open-weights frontier model.

Meta released Llama 3.1 70B with a 128K context window and multilingual support.

Meta released Llama 3.1 8B with a 128K context window and multilingual support.

NVIDIA launched AI Foundry with NeMo, DGX Cloud and NIM services for building and deploying customized Llama 3.1 models.

PSP Investments led a $500 million round with Cisco, Fujitsu, AMD Ventures, EDC, NVIDIA and Salesforce Ventures participating.

CoreWeave announced NVIDIA HGX B200 and GB200 NVL72 Blackwell cluster configurations connected with NVIDIA Quantum-2 InfiniBand.

Mistral AI and NVIDIA jointly released Mistral NeMo, a 12B multilingual open model with a 128K context window.

OpenAI released GPT-4o mini as a lower-cost small multimodal model.

Microsoft Designer exits preview as a standalone app and expands generative design capabilities across Microsoft products.

Anthropic and Menlo Ventures launched a $100 million fund to support early-stage startups building with Anthropic technology.

Hugging Face released compact fully open language models at 135M, 360M, and 1.7B parameter scales.

Mistral AI released Codestral Mamba, an open code model using the Mamba 2 architecture.

Mistral AI released Mathstral 7B, an open model specialized for mathematical reasoning.

Anthropic launched the native Claude application for Android.

Amazon broadly launched its generative-AI shopping assistant in the United States after an earlier beta.

AWS introduced a visual workflow builder for generative-AI applications.

AWS introduced prompt creation, versioning, testing and sharing capabilities.

Perplexity collaborated with Amazon Web Services to launch and scale Enterprise Pro.

Perplexity upgraded Pro Search for more advanced problem-solving, multi-step reasoning and research.

Meta introduced a research pipeline for generating high-quality 3D assets with text and texture control.

Baidu released a faster and more capable version of ERNIE 4.0.

Baidu upgraded its AI coding assistant and repositioned it for enterprise and developer workflows.

DeepSeek updated its chat model with stronger code, math, reasoning and role-play performance.

Amazon hired Adept cofounder David Luan and members of the team while licensing Adept's agent technology.

Perplexity completed a reported financing round of roughly $250 million at a $3 billion valuation, including strategic backing from SoftBank.

Google releases Gemma 2, led by a 27B model with improved performance and efficiency.

Oracle expanded HeatWave from analytics and AutoML into an integrated generative-AI platform.

Oracle released automated in-database vector creation, storage and retrieval for HeatWave.

Oracle released in-database LLMs, automated vector storage and natural-language chat in HeatWave.

Hugging Face refreshed the Open LLM Leaderboard with a harder, contamination-resistant suite and new scoring methodology.

Anthropic expanded Claude availability for US government and intelligence-community customers through AWS.

EvolutionaryScale, founded by former Meta researchers, released ESM-3, a generative protein model originating from Meta's ESM program.

Anthropic launched Projects for organizing chats, knowledge, and collaboration around sustained work.

Investigations alleged that Perplexity accessed publisher content despite robots.txt restrictions; Perplexity disputed aspects of the reporting.

Anthropic introduced Artifacts, a side-by-side workspace for creating and iterating on Claude-generated content.

Anthropic released Claude 3.5 Sonnet with stronger reasoning, coding, writing, and vision performance.

The first Copilot+ PCs become generally available from Microsoft and hardware partners.

DeepSeek released 16B and 236B mixture-of-experts coding models with 128K context.

DeepSeek upgraded its coding API to the Coder V2 generation.

Microsoft releases Florence-2, a unified prompt-based vision model for captioning, grounding, detection and segmentation.

Hugging Face acquired Argilla, an open-source platform for human and synthetic data curation and model feedback.

Microsoft changes Recall from broad Copilot+ PC availability to a Windows Insider preview while strengthening security and privacy protections.