Meta launched Muse Spark 1.1 on July 9. The company priced its first paid AI model far below Anthropic’s Claude and OpenAI’s ChatGPT.The move puts Meta squarely in the coding and agentic AI market, which both rivals currently dominate.

The launch marks a sharp shift from Meta’s open-source Llama strategy. AI chief Alexandr Wang said Meta built the pricing to compete directly with the two market leaders.

Muse Spark’s Aggressive Pricing Play

Meta priced the new model at $1.25 per million input tokens and $4.25 per million output tokens. New accounts get $20 in free credits before billing starts.

Muse Spark 1.1 undercuts all major AI rivals on price. Its $1.25 input rate runs 37% below Sonnet 5’s introductory $2 and 75% below Opus 4.8 and GPT-5.5, which both charge $5.

The output side shows an even wider gap. Meta’s $4.25 rate sits 58% below Sonnet 5’s introductory $10, 83% below Opus 4.8’s $25, and 86% below GPT-5.5’s $30.

That pricing gap could matter most for heavy, high-volume workloads. A developer running the same task across all four models would pay Meta’s rate for less than a third of what GPT-5.5 charges on output alone. The gap also lands amid the broader OpenAI price war with Anthropic.

The gap narrows once Sonnet 5’s introductory pricing expires on August 31, though Meta still comes out cheaper on both ends even against the standard $3/$15 rate.

“Very aggressive and attractive.”
Alexandr Wang, Meta AI chief, to CNBC

Meta detailed the specifics in Meta’s paid API pivot. It marks the company’s first ever charge for AI model access after years of free Llama releases.

Can Muse Spark Overthrow Claude and ChatGPT?

Muse Spark 1.1 reportedly rivals GPT-5.5 and Claude Opus 4.8 on agentic benchmarks, per posts from developers tracking the launch. Meta has not published independent scores against either rival.

Access remains limited. The public preview covers only US developers, requires a waitlist, and stays off third party marketplaces like OpenRouter.

Cheaper access could pull developers toward Meta’s model over time. Whether Muse Spark can match Claude and ChatGPT on real world coding tasks remains unproven.

Источник: BeInCrypto