Meta Launches Muse Wonder and Muse Wonder Full: Benchmarks, Pricing and How to Use It

Meta has released two new agent-oriented models, Muse Wonder and Muse Wonder Full, both supporting a 1M-token context window. Muse Wonder is available through the Meta Model API at a listed price of $4 per million input tokens and $20 per million output tokens, positioning it as a cost-competitive option for long-context agentic workloads.

The larger Muse Wonder Full variant is offered under Muse Code plans and is reported to lead the DeepSWE and SWE-Marathon benchmarks, which evaluate software-engineering agent capabilities. These benchmarks focus on multi-step coding and debugging tasks, suggesting Meta is targeting developers who need models that can reason over large codebases and execute complex workflows.

The 1M-token context window is a notable specification, as it allows the models to process substantially more information in a single pass than many competing offerings. This could be particularly relevant for agent use cases that require ingesting entire repositories, lengthy documentation, or extended conversation histories without truncation.

Meta's entry into the agent-model space with explicit pricing and benchmark claims signals a push toward making agentic AI more accessible to developers through its API infrastructure. Whether these models gain traction will depend on real-world performance beyond the reported benchmark results and how they compare against established competitors in sustained agentic tasks.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.