Meta has published the weights for its new Muse Spark Open model and placed it on Modal’s Auto Endpoints service. The model is designed for agentic coding, long‑horizon tasks, and multimodal input, and it can handle up to one million tokens at once. Modal offers it as a shared endpoint that charges per token. Meta’s internal tests show the model scoring 76.4 on Terminal‑Bench 2.1 and 60 on DeepSWE 1.1.
Why it matters
Developers can now access a large‑context, multimodal coding model through Modal without managing their own infrastructure.
