HomeFeaturesPricingBlogFAQContact
← All articles

Meta Muse Spark 1.1 Pricing: Meta's First Paid AI API Undercuts Claude and GPT

Key takeaways
  • Meta launched Muse Spark 1.1 on July 9, 2026, its first-ever paid AI API, ending a run of free, open-weight Llama releases
  • Muse Spark 1.1 API pricing is $1.25 per million input tokens and $4.25 per million output tokens, with $20 in free credits for new developer accounts
  • Mark Zuckerberg called it "the first time that we're doing a real serious API" and said pricing would be "very aggressive and attractive," per Bloomberg
  • The preview launch is limited to US developers, and Meta is pitching Muse Spark 1.1 as an agentic model built to compete directly with Anthropic and OpenAI on cost</takeaways>

Meta just did something it has never done before: charge developers money for AI. On July 9, 2026, the company unveiled Muse Spark 1.1 pricing alongside the model's API launch, and the headline number is $1.25 per million input tokens and $4.25 per million output tokens, a rate that undercuts several of the leading Anthropic and OpenAI models on the market today.

Why this is a first for Meta

Meta built its entire AI reputation on giving models away. Llama, released open-weight since 2023, was the free alternative to closed labs charging by the token. Muse Spark 1.1 breaks that pattern. According to Bloomberg, this marks the first time Meta has charged businesses for access to one of its models, and Mark Zuckerberg framed it as a deliberate strategic shift, telling Bloomberg ahead of the release, "the first time that we're doing a real serious API," and promising pricing that would be "very aggressive and attractive." The move follows roughly a year after Zuckerberg recruited Alexandr Wang to run Meta Superintelligence Labs, and it signals Meta now wants a piece of the API revenue that Anthropic, OpenAI, and Google have been splitting.

What Muse Spark 1.1 API pricing actually looks like

The numbers are the story here. At $1.25 per million input tokens and $4.25 per million output tokens, Muse Spark 1.1 lands well below Claude Sonnet 5's introductory $2/$10 rate and far below the top tier of OpenAI's newly launched GPT-5.6 family, which prices its flagship Sol tier at $5 input and $30 output per million tokens. Meta is also sweetening the deal with $20 in free API credits for every new developer account, a common tactic to get workloads flowing before the meter really starts running. The preview is US-only for now, so international developers will need to wait for a wider rollout.

The agentic angle

Muse Spark 1.1 isn't being pitched as a general chatbot model, it is positioned as an agentic model, built for the kind of multi-step, tool-using tasks that have become the main battleground among frontier labs this year. Meta is explicitly framing it as a challenger to Anthropic and OpenAI on agentic workloads, not just a cheaper chat completion endpoint. Early coverage from outlets like Cryptobriefing reports Meta claiming Muse Spark 1.1 surpasses some OpenAI and Google models on its internal benchmarks, though as with any vendor-published benchmark, that claim is worth verifying against independent evals once the model sees broader use.

The catch

Aggressive introductory pricing from a company entering a new market is not the same as a durable price advantage. Meta has the balance sheet to run Muse Spark 1.1 at a loss for a long stretch if it decides token-share matters more than near-term margin, but there is no guarantee $1.25/$4.25 survives past the preview window, especially once usage scales past whatever capacity Meta provisioned for launch. It is also a first-generation paid product from a company with far less API operating history than Anthropic, OpenAI, or Google, so reliability, rate limits, and support quality are still unproven at production scale. Developers weighing a switch should treat the launch price as a data point, not a locked-in guarantee, and that kind of side-by-side cost testing across providers is the same instinct behind bring-your-own-key tools like ByteChat, which let you hold keys for multiple providers and compare real output and spend on your own prompts rather than a press release.

Frequently asked questions

Did Meta really start charging for AI access?

Yes. Meta launched Muse Spark 1.1 on July 9, 2026, its first-ever paid AI API, after years of releasing Llama models for free under an open-weight license.

How much does the Muse Spark 1.1 API cost?

Muse Spark 1.1 is priced at $1.25 per million input tokens and $4.25 per million output tokens, with $20 in free credits given to new developer accounts at signup.

Is Muse Spark 1.1 cheaper than Claude and GPT-5.6?

On listed rates, yes. Muse Spark 1.1's $1.25/$4.25 per million tokens undercuts Claude Sonnet 5's introductory $2/$10 rate and is well below GPT-5.6 Sol's $5/$30 tier, though Meta's pricing is a launch preview and could change.

Launch pricing from a company entering a new API market is worth rechecking once the preview window ends and real usage volume kicks in.

Add Meta's models to your side-by-side comparisons

ByteChat is BYOK across every major provider, so when a new model like Muse Spark 1.1 launches you can plug in your own key and compare its output and cost against Claude, GPT, or Gemini on the same prompt, with zero markup on top of the provider's rate.

Try ByteChat free →