Releases

IBM released Granite 4.1 language, vision, speech, embedding and Guardian models under Apache 2.0.

NVIDIA released an open multimodal model unifying vision, audio and language for efficient agents.

OpenAI discontinued the standalone Sora product as its video strategy shifted to newer experiences and models.

DeepSeek introduced the V4 family through Pro and Flash preview API models.

SpaceXAI released its first named Think Fast speech-to-speech voice-agent model.

OpenAI released GPT-5.5 and GPT-5.5 Pro as upgraded frontier models for complex knowledge work and reasoning.

OpenAI released ChatGPT Images 2.0 with stronger world knowledge, instruction following, detail, and an images-with-thinking mode.

Moonshot released Kimi K2.6 with improved agent capabilities and published model weights.

Anthropic released Claude Opus 4.7 with improved software engineering, vision, complex multi-step work, and new cybersecurity safeguards.

Baidu released an open-weight 8B DiT text-to-image model.

Google launches Gemini 3.1 Flash Audio for real-time and batch audio understanding and generation.

Meta Superintelligence Labs introduced Muse Spark, a multimodal model designed for efficient broad intelligence.

Anthropic released Claude Mythos Preview as a gated frontier model for vetted Project Glasswing partners, focusing its advanced cybersecurity capabilities on defensive use.

Microsoft AI launches MAI-Image-2 for faster, higher-quality image generation in Copilot and Microsoft Foundry.

Microsoft AI launches MAI-Transcribe-1 for high-speed multilingual speech-to-text through Microsoft Foundry.

Google releases Gemma 4 open models and makes them available through Google Cloud's AI development stack.

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

Cohere entered speech recognition with the open Transcribe model family.

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

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

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

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

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.

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

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

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.

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

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

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.

Google launches Gemini 3.1 Flash Image for fast, controllable image generation and editing.

Google launches Gemini 3.1 Pro with advances in reasoning, coding and long-running agentic tasks.

Anthropic released Claude Sonnet 4.6 with upgrades across coding, computer use, long-context reasoning, agent planning, knowledge work, and design.

OpenAI released a research preview of GPT-5.3-Codex-Spark, a smaller model optimized for near-real-time coding.

Baidu released a 2.4T-parameter unified model trained jointly across text, image, video and audio.

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.

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

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

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

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

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

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.

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