Zuckerberg launches Muse Ember, Meta’s smallest AI model, at Haiku’s price per token
Meta, led by Mark Zuckerberg, launched Muse Ember, its smallest AI model, priced competitively with rivals like Anthropic's Claude Haiku and OpenAI's GPT-6 Luna. Ember claims higher efficiency per task due to fewer output tokens, according to Meta. It is designed for repetitive tasks within AI agents and is available via Meta's hosted API and cloud partners. Ember's pricing starts at $0.10 per million input tokens and $0.50 per million output tokens, with higher rates for extensive usage.
How this was made
The 30-second read
Why it matters
The announcement may boost Meta's AI narrative but revenue impact remains uncertain.
Market read
Meta's AI model launch adds depth to its AI portfolio, potentially supporting the stock modestly.
What to watch
Meta's lack of downloadable weights and reliance on hosted API could restrict integration, reducing revenue potential
Background
Meta continues its AI push with the Muse family, positioning Ember as a high‑throughput, low‑cost sub‑agent model.
Ticker impact
Meta announced its new Muse Ember AI model with detailed token pricing and efficiency claims, marking the first public disclosure of the product.
likely modest upside as investors price in Meta's expanded AI services
While the model itself may not drive immediate revenue, the lower per‑task cost could attract developers and bolster Meta's AI strategy, supporting a slight price lift.
Market effects
strengthens the AI/cloud services sector by adding a low‑cost competitor to Anthropic and OpenAI
U.S. tech market may see modest gains as AI spend shifts toward cost‑effective models
global AI developers could adopt the model, modestly influencing worldwide AI infrastructure spending
Counterpoint
The model's limited availability and higher tier pricing may limit adoption, dampening any upside
Key entities
- companyMeta Platforms
U.S.-listed tech giant launching Muse Ember
- personAlexandr Wang
Head of Meta Superintelligence Labs, quoted on Ember



