Meta's open-weight reasoning model lands on AI Gateway with a 1M-token window for code and screenshots
Synopsis
Meta's open-weight reasoning model Muse Spark Open is now available on AI Gateway, aimed at coding and agentic workflows such as multi-file refactoring, debugging, and understanding large codebases. It accepts text and image input with a 1M-token context window, letting agents work with code and screenshots in the same request. Users call it as meta/muse-spark-open across the AI SDK, the OpenAI-compatible Chat Completions and Responses APIs, the Anthropic-compatible Messages API, and coding agents connected to AI Gateway, without creating separate provider accounts.
Interpretation
Coding agents can now receive code and screenshots in the same request instead of converting image content to text first. Models wired into coding agents have generally handled text input only, with image information requiring separate conversion or a separate call. Muse Spark Open supports text and image input with a 1M-token context window, and the announcement describes its purpose as letting agents work with code and screenshots in the same request.
Calling the model does not require a separate Meta provider account, and the model name stays the same across API formats and coding agents. Adding a new model has typically meant configuring accounts and interfaces per provider. The model is used as meta/muse-spark-open across the AI SDK, the OpenAI-compatible Chat Completions and Responses APIs, the Anthropic-compatible Messages API, and coding agents connected to AI Gateway, with ZDR and no training.
Enabling it in Claude Code, Codex, Cursor, and similar agents goes through AI Gateway's coding-agents setup flow. Switching models inside these agents previously required separate configuration for each. After installing the latest Vercel CLI, running vercel ai-gateway coding-agents setup and then selecting meta/muse-spark-open in the agent.
Perspective
The model targets coding and agentic workflows, including multi-file refactoring, debugging, and understanding large codebases, and supports text and image input with a 1M-token context window. Access covers the AI SDK, the OpenAI-compatible Chat Completions and Responses APIs, the Anthropic-compatible Messages API, and coding agents connected through AI Gateway, with ZDR and no training.
The source is a launch announcement and provides no benchmark results or capability comparison with other models, so the model's actual performance on coding and agentic tasks remains unclear. The 1M-token context window is a model specification, and the announcement does not describe practical usability or cost behavior at that length.