Meta launches Muse Spark 1.1, a paid AI coding model, as it races to catch OpenAI and Anthropic

What's new

Meta launched Muse Spark 1.1 on Thursday, a multimodal AI model for coding and agentic tasks. It is available via a new API, marking Meta's first paid offering. The model handles complex code migrations, bug fixes, and multi-step workflows. Meta describes it as a "step-change" from its April predecessor.

Muse Spark 1.1 delivers exceptional performance in personal agentic tasks that require planning and orchestration across a range of external apps and services.

Paid access and pricing

For the first time, Meta is charging for API access. It costs $1.25 per million input tokens and $4.25 per million output tokens. That is competitive with Anthropic and OpenAI, and significantly cheaper than high-end models like Anthropic's Fable. The move opens a new revenue stream amid massive AI investments.

Competitive landscape

Muse Spark 1.1 was developed under Chief AI Officer Alexandr Wang. On benchmarks, it performs near the levels of models from OpenAI, Anthropic, and Google. Meta is racing to catch up, with a more powerful model called Watermelon slated for later this year.

A strong agentic and coding model at a very low price.

These were priority areas for us.

A busy week for AI

The launch was part of a flurry of AI announcements this week. Meta also debuted Muse Image on Tuesday, while xAI and OpenAI released new models on Wednesday and Thursday. Zuckerberg promoted Spark on X, his first post in three years.

What analysts are watching

The shift to a paid API and the promise of upcoming models signal Meta's serious intent in the AI arena. Wang envisions future agentic AI handling tasks like planning vacations. With billions of users and data, Meta's strength may lie in integration with its apps.