Meta released Muse Spark 1.1 on Thursday, opening a public API preview for its coding-focused AI model and publishing pricing for the first time. AI chief Alexandr Wang told CNBC the model represents Meta’s “strongest model for agentic and coding work yet,” positioning it as a direct competitor to Anthropic’s Claude Code and OpenAI’s Codex.
The pricing is notable: $1.25 per million input tokens and $4.25 per million output tokens, with $20 in free credits for every new API account. Wang characterized the rates as “very aggressive and attractive” compared to Anthropic and OpenAI offerings.
From Private Preview to Public Access
The original Muse Spark launched in April under Wang’s leadership, available only to “select partners” through a private API preview. Muse Spark 1.1 opens access through a developer portal where users can sign up for integration instructions, according to CNBC. A Meta spokesperson said early partners already have API access, with new users joining through a waitlist.
Meta is limiting API access to its own properties for now, keeping Muse Spark off third-party platforms like OpenRouter. “This is going to be served on top of the computer infrastructure that we’ve built,” Wang said.
The update is Meta’s second Muse family rollout this week. On Tuesday, Meta released Muse Image (originally code-named Mango), a model for image generation aimed at creators and advertisers.
The Revenue Pressure Behind the Launch
Meta CEO Mark Zuckerberg faces growing Wall Street pressure to demonstrate returns on the company’s massive AI infrastructure investment. Meta spends at the rate of its hyperscaler peers but lacks a cloud infrastructure business, though it plans to start one, according to CNBC. The company has also struggled to keep pace with OpenAI, Anthropic, and Google in developing popular models and AI applications.
Muse Spark 1.1 represents Meta’s first paid API product, a meaningful shift for a company that built its AI reputation on open-source releases. The pricing signals that Meta views the agentic coding market as a revenue opportunity, not just a research contribution.
The Competitive Landscape
Every major AI lab now ships agent-optimized model variants. OpenAI has Codex. Anthropic has Claude Code. Google has Gemini’s coding capabilities. Meta’s entry with aggressive pricing and a dedicated API puts a fourth well-funded competitor on the field. The pricing undercut matters most for teams running high-volume agent workflows, where token costs compound across multi-step reasoning loops.
The constraint is distribution. By limiting Muse Spark 1.1 to Meta’s own infrastructure and keeping it off third-party marketplaces, Meta is betting that developers will come directly rather than discovering the model through aggregators. That trade-off gives Meta more control over the inference economics but limits the model’s reach during a critical adoption window.