Meta releases its first frontier model, Muse Wonder
- ID
- 34596
- Status
- summarized
- Published
- 13 Oct 2026, 1:56 AM
- Fetched
- 13 Oct 2026, 2:42 AM
- Provider
- TechCrunch
- Category
- technology
- Original URL
- https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/
- Source URL
- https://techcrunch.com/feed/
Summary
- Score
- 5.0
- Created
- 13 Oct 2026, 2:42 AM
- Tags
- Audience
- developersai_ml_learnersai_agent_userssaas_founders
What happened
Meta's Superintelligence Lab released Muse Wonder and Muse Wonder Full, described as its first frontier models and larger, more capable successors to Muse Spark with better coding results. Meta's own model card claims benchmark scores comparable to Fable 5.1 and GPT-6 Astra, with particular emphasis on "active memory" and a MemoryBench score Meta says beats Fable, Opus, and Astra. Wonder is priced at $4 per million input tokens and $20 per million output tokens via the Meta Model API, roughly equal to GPT-6 Sol or Opus 5.5, while Wonder Full is limited to Muse Code subscribers and is not served through the API.
Why it matters
The pricing is the actionable part: at $4/$20 per million tokens, Wonder is priced at parity with GPT-6 Sol and Opus 5.5, so there is no cost reason to switch providers — the differentiator Meta is selling is memory retention, not price. Also decide deliberately about the API split: if Wonder Full is the stronger model and is only reachable through a Muse Code subscription, then any product you build on the API tops out below Meta's best model, which matters if you were planning to route agent workloads there.
Discussion angle
Meta is holding its strongest model (Wonder Full) off the API and behind a Muse Code subscription while claiming MemoryBench leadership — ask whether you'd build a product on a model family whose best tier you can't call programmatically, and how much weight to give self-reported benchmark numbers versus the pricing parity with GPT-6 Sol and Opus 5.5.