Meta has announced the launch of Muse Spark, a new large language model (LLM) built in house for the social media firm’s products.

The tech giant said it is the first in a new line of Muse LLMs developed by its Meta Superintelligence Labs division and will serve as a small-scale base on which future models can be built.

The model now powers the Meta AI assistant in its own app and website. The company says it can handle complex problems in “Thinking mode” by running multiple subagents in parallel to solve discrete parts of a long prompt at the same time, speeding up responses.

Muse Spark can also respond to visual prompts, which Meta describes as ‘multimodal perception’. As an example, the company said that the tool can rank a shelf of snacks by protein content or estimate the calories in a plate of food from a photo. Meta said this feature helps Muse Spark to offer detailed responses to health questions, adding that it was developed in collaboration with a team of physicians.

The tool can also build websites and mini-games from simple prompts.

These features are now standard offerings for flagship models, available in Anthropic’s Claude, OpenAI’s ChatGPT, and Mistral’s Le Chat. However, Meta’s model is also plugged into the company’s social media apps.

Its shopping mode can source products based on “the styling inspiration and brand storytelling” taking place on Instagram, Facebook and Threads. The model can also aggregate posts to enhance its recommendation ability for instance by finding coffee shops in an area with large amounts of positive posts.

The upgraded modes, Thinking and Instant, are currently available on the Meta AI app and meta.ai website, while the new features are in the process of being rolled out in the US. Meta said that it will bring the features to more countries and services over the coming weeks, including WhatsApp, Instagram and its AI glasses.

---