Meta Unveils Frontier AI Models Muse Wonder and Muse Wonder Full

Meta’s Superintelligence Lab has introduced its first frontier‑class language models, Muse Wonder and Muse Wonder Full, positioning them as direct competitors to the latest offerings from OpenAI and Anthropic.

Performance and Pricing

According to Meta’s internal benchmarks, the Wonder series matches or exceeds the scores of Fable 5.1 and GPT‑6 Astra on standard language and coding tasks. Access via the Meta Model API is priced at $4.00 per million input tokens and $20.00 per million output tokens, putting it in the same cost bracket as GPT‑6 Sol or Opus 5.5.

Memory Capabilities

A key focus of the release is “active memory,” a feature designed to let the model retain and synthesize user‑provided information over extended periods. Meta reports that Muse Wonder now leads the MemoryBench benchmark, outperforming Fable, Opus and Astra in continual learning scenarios.

Why it matters for GPU / AI infrastructure: The training of these models was carried out at Meta’s newly operational Prometheus data center in Ohio, highlighting the growing demand for high‑density GPU clusters capable of handling frontier‑scale workloads. Enterprises evaluating AI infrastructure should note that similar performance targets will require comparable compute footprints and memory bandwidth.