Cursor adds Meta's Muse Wonder with a 46.1% CursorBench score

Cursor has added Muse Wonder, a model from Meta Superintelligence Labs, to its coding environment. Cursor's published evaluation lists a 46.1% score at the model's High reasoning setting, which is the default configuration.

Pricing is $4 for each million input tokens and $20 for each million output tokens. Cursor supports a standard context capacity of 300,000 tokens, expanding to 1 million in Max Mode without increasing the per-token rates. The model also accepts images.

Meta introduced Muse Wonder and Muse Wonder Full on October 12 for agent tasks spanning personal and professional work. At launch, Wonder was accessible through Meta's API and Muse Code, while the Full version was limited to Muse Code subscribers.

Inside Cursor, Wonder can use tools to inspect and edit files, execute terminal commands and operate a browser. Some Teams and Enterprise administrators must enable Meta in their provider settings before members can select the model.