Model launchesOpenAI releases GPT-5.6 family — Sol, Terra and Luna optimize efficiency, cost and professional task performanceOpenAI has launched the GPT-5.6 model family — Sol (flagship), Terra (balanced), and Luna (cost-efficient) — for general availability after a limited preview.6 min read
Model launchesSensorFM: a large foundation model trained on over a trillion minutes of wearable sensor dataSensorFM is a large sensor foundation model pre-trained on more than one trillion minutes of multimodal wearable data from five million consented participants (Sept 2024–Sept 2025).5 min read
Model launchesMeta launches Muse Spark 1.1: a multimodal agentic model with million-token contextMeta Superintelligence Labs released Muse Spark 1.1, a multimodal reasoning model optimized for agentic tasks, available now in the Meta AI app and via the public-preview Meta Model API.4 min read
Model launchesOpenAI launches GPT-Live-1 to improve conversational flow and reduce interruptionsOpenAI introduced GPT-Live-1, a conversational model designed to listen more naturally, wait through pauses, and interrupt less during exchanges.2 min read
Model launchesOpenAI launches GPT‑Live voice mode that delegates hard tasks to GPT‑5.5OpenAI has upgraded ChatGPT's voice mode to a new GPT‑Live model that can delegate web searches, deeper reasoning, and complex tasks to GPT‑5.5 running in the background.3 min read
Model launchesSpaceX launches Grok 4.5 — a cost-focused coding and agents model trained with CursorSpaceX on Wednesday released Grok 4.5, its first AI model trained specifically for coding and autonomous agents, developed in partnership with Cursor following a $60 billion acquisition.8 min read
Model launchesOpenAI launches GPT-Live to enable continuous, conversational voice with separate speech and reasoning layersOpenAI introduced GPT-Live, two full‑duplex voice models (GPT‑Live‑1 and GPT‑Live‑1 mini) that let the assistant listen and speak simultaneously and separate the real‑time voice layer from a background reasoning model (initially GPT‑5.5).7 min read
Model launchesGrok 4.5 and Composer different model-weight classes; Composer 2.5 remains availableAccording to the provider's announcement, Grok 4.5 and Composer are different model weight classes; the Composer 2.5 version will remain on offer, and in the future new models of the same 2.5 size will be released.1 min read
Model launchesGrok 4.5 co-developed with SpaceXAI for uses beyond software developmentThe announcing company trained the Grok 4.5 model in partnership with SpaceXAI; according to the new announcement, this is the company's most powerful model to date (Grok 4.5) and the first designed for broader applications beyond software development.1 min read
Model launchesOpenAI introduces GPT-Live voice models, available in ChatGPTOpenAI introduced a new generation of voice models called GPT-Live, aimed at natural human–AI voice communication; the feature is being rolled out in ChatGPT starting today.1 min read
Model launchesGPT-Live: the interactive, speech-optimized AI voice modelGPT-Live is a new speech-based artificial intelligence voice model designed to make communication sound more human and continuous; according to the developers, it is their most intelligent voice model.1 min read
