A foundation model is trained on broad datasets at large scale so it can serve as a base for many downstream applications. Rather than learning only one narrow task, it develops general capabilities that can be accessed through prompting, connected tools, retrieval, or additional training. Large language models are a familiar example, while other foundation models work with images, audio, video, code, or several data types together. Building them requires extensive computing resources and careful data preparation, but many organizations use existing models through hosted services or open weights. Their broad reuse also spreads underlying limitations: bias, security weaknesses, licensing concerns, and factual errors may affect numerous products unless developers evaluate the model within each specific context.
HappyHorse-1.0, an open-source AI video model, has topped global benchmarks, outperforming leading proprietary systems and signaling a shift in the video generation market.
Meta shares climbed 7% after unveiling Muse Spark, its new multimodal AI model, signaling investor confidence in the company’s revamped AI strategy.
Meta has unveiled Muse Spark, a multimodal AI model with advanced reasoning and multi-agent capabilities, marking a step toward its vision of personal superintelligence.
Microsoft has introduced MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2, expanding its AI model lineup with faster performance and competitive pricing for developers. The models are now available through Microsoft Foundry.
SoftBank has secured a $40 billion bridge loan to deepen its investment in OpenAI and accelerate its broader AI strategy.
AI-driven bots now generate more internet traffic than humans, according to a new report, highlighting the growing impact of automated systems online.
New research shows that telling AI models they are “experts” can reduce accuracy in tasks like coding and math, while improving safety and alignment outcomes.
Encyclopaedia Britannica and Merriam-Webster have sued OpenAI, alleging the company copied nearly 100,000 articles and dictionary entries to train ChatGPT without permission.
Meta introduced four new in-house MTIA chips designed for AI training and inference as the company accelerates data center expansion. The chips aim to improve performance and reduce reliance on external hardware suppliers.
Google has rolled out Canvas in AI Mode to all U.S. users in English, enabling people to create documents, dashboards, and interactive tools directly within Search. The feature provides a dynamic workspace for planning projects and building simple applications.