A New Competitor in AI Changes the Price for Everyone
Meta just entered the frontier AI race, and what matters most is how that pressure will bring down prices for everyone building with AI.
- $4
- per million input tokens
- $20
- per million output tokens
- $145B
- Meta's 2026 AI infrastructure budget
- Oct 12, 2026
- Muse Wonder launch date
Meta released Muse Wonder on October 12 — its first model it says can match the best AI from OpenAI and Anthropic. For anyone paying to use AI right now, this matters. In markets, when a giant enters, prices move.
Muse Wonder costs $4 for every million input tokens — the small chunks of text an AI reads — and $20 for every million output tokens. That puts it at the same price as Anthropic's Claude Fable 5.1 and OpenAI's GPT-6 Astra. More players at the top means more pressure on price.
Meta spent $145 billion on AI infrastructure in 2026. That number alone tells you this is not a short-term experiment. The models also come with 'Active Memory,' which lets the AI recall things you have told it in earlier conversations, and 'Active Focus,' which helps it try different approaches when a task hits a wall.
The models are aimed at consumers first — not at companies managing contracts or workflows. But the ripple effect on business is real. More competitors at the frontier push the others to lower prices or get better — and usually both happen.
If you choose or pay for an AI model today, this is a good time to compare options and watch for price drops. Frontier AI is getting more crowded. What costs $20 per million output tokens today will likely cost less next year.