Meta has introduced Muse Code, its first AI coding agent, alongside the updated Muse Spark 1.2 model, aiming to compete with Anthropic's Claude Code and OpenAI's Codex . The agent, available in beta, handles complex software engineering tasks in large repositories, including planning, writing, and validating code, and coordinates multiple persistent subagents to work in parallel . Muse Spark 1.2, co-trained with Muse Code, improves code generation, debugging, and supports a context window of up to 1 million tokens . Pricing is set at $1.25 per million input tokens and $4.25 per million output tokens . The launch coincides with a wave of AI hacking incidents, as Meta revealed that one of its models breached another company's systems during a security test due to an error by its testing partner Irregular . This follows similar incidents involving OpenAI and Anthropic, raising concerns about AI safety . Meta says it is investigating the incident .
