Meta Expands Muse Agent Capabilities with New LLMs
Meta has introduced two new large language models, Muse Wonder and Muse Wonder Full, as part of its strategy to strengthen the Muse agent ecosystem. The company claims these models are competitive with the latest from OpenAI and Anthropic, based on internal benchmarks.
Access and Integration
- Muse Wonder is available to developers via the Meta Model API, while Wonder Full is restricted to Muse Code subscribers.
- Both models are set to power the Muse agent, with Meta planning to develop additional specialized models for broader use in the future.
Performance Claims
Meta asserts that Wonder and Wonder Full match the performance of leading models such as GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1. However, no independent or third-party benchmarks have been released to substantiate these claims, and there is no comparative data on Google’s Gemini 4 Argon.
Implications for Indie Builders
For solo operators and small teams invested in the Muse agent ecosystem, these new models may offer improved capabilities and more competitive options for high-end applications. However, it remains critical to validate performance in your own workflows before making significant changes.
Caveat
All reported performance metrics are from Meta’s internal testing. Wait for independent benchmarks and user feedback before committing to a migration or integration.