THE HOOK
Meta has stopped shipping lightweight warm-ups and finally put a frontier model on the table.
THE GIST
Meta's Superintelligence Lab released Muse Wonder and Muse Wonder Full, trained at its gigawatt-scale Prometheus data centre. Wonder costs $4 per million input tokens and $20 per million output tokens through the Meta Model API, while Wonder Full is reserved for Muse Code subscribers.
THE TAKE
Pricing it level with the top labs is a statement of confidence, and a risky one. Developers don't switch on benchmark parity alone; they switch when something is cheaper or clearly better. Keeping the best version behind a Muse Code subscription tells you where Meta really wants your money. Bold, but not open.
OPINION
TechCrunch notes Meta's own model card puts Wonder level with Fable 5.1 and GPT-6 Astra, with a lead on the MemoryBench memory test, which is exactly the homework you'd expect a newcomer to mark itself. The full safety report is still to come, so the CBRN and cyber-misuse claims are Meta's word for now.
- Story type: Important story
- Impact: High
- Confidence: High (official announcement)
- Audience: Developers
- What changed: Meta released Muse Wonder and Muse Wonder Full, its first frontier models, with Wonder priced at $4/$20 per million tokens on its API.
- What to watch: Meta's full safety report and whether independent benchmarks back up its parity claims against Fable 5.1 and GPT-6 Astra.
- Reading time: 1 minute
