Meta Pushes Into Commercial Frontier APIs With 1M-Context Muse Spark 1.1 and Muse Ember
Meta Superintelligence Labs prices its closed 1-million-context Muse Spark 1.1 model at $1.25 input tokens while shipping Muse Ember as a $0.10 agent execution tier.

By Jin, Founder & Startups
, updated 2 min read
Meta Superintelligence Labs is pressing deeper into direct API revenue with Muse Spark 1.1 and the November 9 launch of lightweight execution model Muse Ember, offering closed-weight multimodal reasoning across a 1,048,576-token context window.
In its Muse Spark 1.1 launch post, Meta positioned the closed Meta Model API directly against Anthropic and OpenAI on agentic execution cost, tool orchestration, and computer use.
Two-Tier Commercial Strategy for Agent Orchestration
Muse Spark 1.1 flagship API: Supports native text, image, and video inputs alongside active context compaction at $1.25 per million input tokens ($0.15 cached) and $4.25 per million output tokens, plus $20 in US preview credits and a deeply discounted contributor option for developers permitting training opt-in.
Muse Ember volume executor: Released November 9 with 1M input and 512K output capacity, native audio/video/PDF support, and $0.10 per million input tokens up to 200K context ($0.30 above, $0.50 output, $0.004 cached), reaching 36 on Artificial Analysis's Intelligence Index at roughly one-third of comparable per-task agent cost.
Cloud distribution: Alongside Meta's Muse Code IDE and internal API, Meta placed Muse Ember on Google Cloud Vertex AI and Oracle Cloud to capture enterprise developer spend outside open-weight Llama downloads.
Sources
- 1Meta AI Blog: Introducing Muse Spark 1.1ai.meta.com