Meta AI

DINOv3 scales self-supervised learning (SSL) for images to produce universal vision backbones

SAM 2 is a segmentation model that enables fast, precise selection of any object in any video or image.

DINOv3 scales self-supervised learning (SSL) for images to produce Meta's strongest universal vision backbones.

V-JEPA 2 is described as the first world model trained on video that achieves state-of-the-art visual understanding and prediction.

Movie Gen is described as the most advanced family of media foundation AI models empowering immersive storytelling.

Meta AI conducts work across AI Infrastructure, Generative AI, NLP, Computer Vision and other core areas of AI.

Meta AI's responsible AI core principles are Privacy & Security, Fairness & Inclusion, Robustness & Safety, Transparency & Control, and Accountability & Governance.

Muse Spark 1.1

Multimodal reasoning model built for agentic tasks with major gains in tool and computer use, coding, and multimodal understanding

1 million token context window with active context management

Public preview of the new Meta Model API launched for developers to access Muse Spark 1.1

Available in "Thinking" mode in the Meta AI app and on meta.ai

Developed by Meta Superintelligence Labs

Muse Image is Meta's most advanced image generation model that follows instructions faithfully, edits with precision, and composes from multiple references

Muse Video delivers exceptional visual fidelity with native audio support and is built on the same pretraining base as Muse Image

Muse Image and Muse Video are the first media generation models developed by Meta Superintelligence Labs

Muse Image operates as an agent that invokes search and coding tools to improve accuracy, self-refines its own generations, and improves through scaling test-time compute

Muse Image integrates with Muse Spark, allowing the two models to share tools and plan jointly for agentic media generation

Muse Image is available on the Meta AI app, meta.ai, Instagram Stories in the US, and WhatsApp in limited countries, and is coming soon to Facebook; Muse Video is coming soon to creators and Meta AI

Brain2Qwerty uses AI to decode brain activity into text without any surgical implant (non-invasive).

Brain2Qwerty v2 is an end-to-end deep learning pipeline that decodes sentences from raw brain signals in real time.

Brain2Qwerty v2 achieves a 61% word accuracy rate, improving on the 8% word accuracy of other non-invasive methods.

For the best participant, Brain2Qwerty v2 achieves 78% word accuracy, with more than half of all sentences decoded with one word error or less.

Brain2Qwerty v2 was trained on data captured from participants typing while wearing a magnetoencephalography (MEG) device.

Meta is releasing the full training code for Brain2Qwerty v1 and v2 to accelerate neuroscience research.