Meta's Muse Wonder Costs Exactly What Opus 5.5 Costs. It Loses 9 of Meta's Own 14 Benchmarks

Meta released Muse Wonder and Muse Wonder Full on October 12, 2026, positioning them as their most capable models with state-of-the-art performance on personal and professional agentic tasks. Muse Wonder is available on the Meta Model API as muse-wonder-1.0 and on OpenRouter as meta/muse-wonder, while Muse Wonder Full is exclusive to Muse Code subscribers and not served through the API. The full model card lists a 1M-token context and supports text, image, video, and audio inputs with text outputs.

The article compares Muse Wonder Full against Claude Opus 5.5 and GPT-6 Astra, and Muse Wonder against these models and GPT-6 Astra on 14 benchmarks. The API model, Muse Wonder, costs $4 per million input tokens and $20 per million output, while Muse Wonder Full is only available to Muse Code subscribers. The benchmark results show that Muse Wonder, the API model, loses to Opus 5.5 in 9 out of 14 benchmarks but wins in 5 benchmarks, specifically in Legal Agent Work (Harvey LAB-AA), GDP.pdf, Memory Bench (MemoryBench), and two long-horizon coding tasks, DeepSWE 1.1 and ProgramBench.