Meta launches AI coding model to compete with OpenAI, Anthropic

Meta publicly released Muse Spark 1.1 on Thursday, a multimodal AI model designed for agentic coding. The company is aiming to compete with similar offerings from major players like OpenAI and Anthropic.

The new model, Spark 1.1, is capable of multi-step reasoning and managing complex processes, including digital workflows and deploying new features in enterprise systems. Meta enters this space somewhat later than its competitors, as both Anthropic and OpenAI have offered comparable models for some time.

Cost of usage is a significant factor in the AI industry, and Meta appears to be offering competitive rates. The company plans to charge $1.25 per million input tokens and $4.25 per million output tokens, placing it in proximity to Anthropic's Claude Haiku and OpenAI's GPT-5.6 Luna.

Meta's primary pitch for Spark 1.1 focuses on its ability to handle large agentic workloads, fix bugs, and assist with code migrations—automation tasks that enterprises are increasingly seeking. The launch was notable enough that Meta CEO Mark Zuckerberg posted about it on X, his first post in three years.