Meta has made an unexpected yet highly aggressive move by releasing a commercial version of its language model, Muse Spark 1.1. The company, which previously focused exclusively on open-source solutions like Llama, is entering the paid AI services market for the first time. And it is doing so with a price point that could significantly reshape the balance of power in the industry.
The cost of using the Muse Spark API is set at $1.25 per 1 million input tokens and $4.25 per 1 million output tokens. These figures appear to be a direct challenge to market leaders. For comparison, processing the same volume of input data costs $2 for Anthropic Sonnet 5, and $5 each for flagship models from OpenAI GPT-5.5 and Anthropic Opus 4.8. The gap is enormous: savings for developers range from 37% to 75%.
Price Dumping as a Strategy
The difference in text generation costs is particularly striking. $4.25 for Meta versus $10 for Sonnet 5, $25 for Opus 4.8, and $30 for GPT-5.5. This is not just a discount, but a deliberate disruption of the current pricing model. Meta is evidently willing to operate with minimal margins to capture market share where direct competitors currently dominate.
According to available data, Muse Spark 1.1 demonstrates performance comparable to GPT-5.5 and Claude Opus 4.8 in complex tasks involving autonomous code execution and AI agent operations. However, there are no independent benchmarks yet, and access to the platform is limited—only developers in the United States can apply for testing.
My analysis shows: Meta is launching not just a model, but a full-scale price war. If the quality of Muse Spark is confirmed in practice, it will deal a serious blow to the positions of OpenAI and Anthropic, especially in the enterprise development segment, where every dollar counts.