Meta Unveils LLM Muse Wonder for Coding and Agents

Meta has unveiled two new flagship language models, Muse Wonder and Muse Wonder Full, designed for programming tasks, autonomous agent creation, and computer control.

The models feature a one-million-token context window, active memory for information retention between sessions, and goal-holding mechanisms for long-running tasks. Sub-agent parallel execution is also supported.

API pricing for Muse Wonder is $4 per million input tokens and $20 for output. Cached input costs just $0.20, while using long contexts incurs no additional fees.

Muse Wonder Full is currently unavailable via API; it is exclusively accessible through the Muse Code subscription service, priced between $50 and $200 monthly. According to Meta's tests, Wonder outperforms Astra and Opus.

By comparison, GPT-6.1 Sol is twice as cheap at $2/$10 per million tokens, positioning itself as a coding model approaching Astra's quality. Notably, the company previously associated with open-weight Llama now employs a closed distribution strategy.