Models News & Updates

Agentic Vision, a new capability in Gemini 3 Flash, combines visual reasoning with code execution to improve image understanding and accuracy.

Arcee AI releases Trinity, a 400B-parameter foundation model, as a permanently open alternative to existing models, with plans for multi-modal support and competitive API pricing.

Nvidia releases Earth-2 weather forecasting models, promising faster and more accurate forecasts.

OpenAI retires GPT-4o, citing a focus on improving more widely used models. Users who have been preferring 4o's conversational tone will experience notable changes. Although conversations with 4o will not be deleted, the model will no longer be accessible through ChatGPT, except via API and for users who were already using it.

Google DeepMind has released Project Genie, an experimental research prototype that lets users create, explore, and remix interactive worlds, to Google AI Ultra subscribers in the U.S.

Grok Imagine is a new multimodal release that brings five new model endpoints to a single creative stack, covering generation and editing across image and video workflows.

Moonshot introduced Kimi K2.5, an open-source AI model that can generate web interfaces from images or video and supports coding capabilities, marking a significant step forward for 'vibe coding' tools.