Meta officially launched Muse Spark 1.1, a new multimodal AI model designed for agentic coding. This marks Meta's entry into the competitive AI coding market, aiming to rival offerings from OpenAI and Anthropic. The company highlights Spark's ability to handle large agentic workloads, fix bugs, and assist with major code migrations, catering to enterprises seeking AI automation.
Meta is positioning Muse Spark 1.1 with an aggressive pricing strategy. It will charge $1.25 per million input tokens and $4.25 per million output tokens, making it significantly cheaper than some competitors. For comparison, OpenAI's GPT-5.5 charges $5 per million input tokens and $30 per million output tokens, while Anthropic's Claude Opus 4.8 costs $5 and $25, respectively. Google's Gemini 3.1 Pro is priced at $2 per million input tokens and $12 per million output tokens. Meta is also providing $20 in free credits for new API accounts.
Analysts view Meta's aggressive pricing as a potential game-changer, forcing rivals to consider price adjustments. Pareekh Jain of Pareekh Consulting noted that the low output token price, about 86% below GPT-5.5 and over 90% below Claude Opus 4.8, is particularly attractive for coding, customer service, and process automation agents where output tokens are a major expense. However, some analysts, like Jena, suggest that Meta might increase prices after establishing market share, similar to past patterns in advertising and cloud services.
Despite the competitive pricing, experts emphasize that quality and enterprise-grade features like governance, security, reliability, and support will remain crucial differentiators. Muskan Bandta of ZopDev highlighted that while price attracts initial interest, capabilities retain users. Nonetheless, Meta's move is expected to intensify competition, potentially leading to lower costs across the industry and encouraging multi-model procurement among enterprises.
