Meta Opens Muse Spark 1.1 to Developers — and Bets on Aggressive Pricing

Meta updated its large language model Muse Spark on Thursday and is making it available to outside developers for the first time. The new Muse Spark 1.1 is available now as a public API preview for US developers through the new Meta Model API. It marks the first time Meta has charged businesses for access to its models — opening up a new revenue stream in the process.

What’s new

Meta calls Muse Spark 1.1 a “step-change” over the first generation, which launched in April. The improvements focus in particular on coding — including detecting and fixing complex bugs — as well as longer and multi-step tasks. Meta also cites better support for end-to-end agentic workflows across multiple apps, including multi-agent systems, along with native multimodal perception across images, videos, and documents.

The model already powers a wide range of features across Facebook, Instagram, WhatsApp, and the Meta AI app. Version 1.1 is available now in “Thinking” mode through the Meta AI app and website. Meta is including $20 in free credits with every new Meta Model API account.

Alexandr Wang, who was brought in to reboot Meta’s AI effort and under whom Muse Spark was developed, described the gains in coding and agentic tasks as the key focus of this release: “These were priority areas for us.”

The price war with OpenAI and others

Pricing is at the heart of the announcement. In an interview ahead of the release, Mark Zuckerberg promised “aggressive” pricing — in a crowded market for AI tools, Meta wants to win on price. The company charges $1.25 per million input tokens and $4.25 per million output tokens.

That positions Muse Spark 1.1 below several competing offerings: it is cheaper than SpaceXAI’s new Grok 4.5 and also undercuts Anthropic’s Opus model. Meta itself frames the price as attractive given the model’s capabilities and calls it one of the most affordable options on the market.

The move comes during an especially busy stretch of model updates. In this week alone, OpenAI and SpaceXAI also debuted faster models.

Context

With the move, Meta is trying to close the gap with AI leaders OpenAI, Google, and Anthropic — following a series of high-profile hires and a company restructuring last year. Just earlier this week, Meta unveiled Muse Image, an image-generation model that powers new editing features in Instagram and has proved controversial for incorporating other users’ content.

Muse Spark 1.1 is explicitly not the big leap Meta is aiming for. A significantly more powerful model code-named “Watermelon” is still in training, uses vastly more computing power, and is expected to be released later this year.

Longer term, Wang is pursuing a vision of a more agentic Meta AI that could handle tasks like planning a party or a vacation and take more direct action on its own. He sees Meta’s strength in its billions of users and how much the company knows about them: “That’s really the part that no one else can replicate.”