Muse Spark Open

What you can use today

Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.

Exact catalog ID: meta/muse-spark-open. Reads text, images, audio and video and produces text. Tool support lets an application connect it to actions such as searching files or running code; your application supplies those tools.

Available thinking settings: low, medium, high, xhigh. We do not yet have comparable independent results for these settings. The documented self-hosted model default is xhigh when effort is omitted. A default is not a recommendation for every task.

This model is not served by Meta Model API. Choose a hosting partner or download it under the Muse Model License. The compressed weights alone occupy about 537 GB; full-precision serving needs roughly 2 TB of GPU memory. The 42B active parameter count does not mean the whole model fits on a small GPU.

The model supports 1,048,576 combined context tokens. The OpenRouter offer shown here limits generated output to 943,718 tokens; other hosts can set different limits.

Rates describe the catalog offer observed on 2026-11-03. Allow for retries and any tool fees. Models that think before answering may also charge for that work.

Provider release and specifications ↗

Inspect the exact catalog model ↗

Test it before switching

Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. A cheaper token price does not guarantee a cheaper completed task.