Muse Spark 1.3 Contributor API

It may. Muse Spark 1.3 Contributor is Meta's Contributor tier, and Meta's docs say that content sent to it, meaning your prompts, code, files and context and the answers you get back, may be used to train future Meta models, in exchange for the lower price. At Meta's standard tier, content is not used to train Meta models. So don't send this model anything private, confidential or personal, or anything you have promised others to keep private. For that work, use Muse Wonder (muse-wonder-1.0), which runs at Meta's standard tier.

On Zurelay, muse-spark-1.3-contributor costs $0.04 per 1M input tokens and $0.08 per 1M output tokens, with cached input at $0.001. Meta's Contributor-tier price is $0.10 input and $0.20 output per 1M, so you save 60%. The rate is the same across the whole context window, and reasoning tokens bill as output.

Meta runs the same model at both tiers. The difference is data use and price: Meta may train future models on what you send to the Contributor tier and doesn't at the standard tier, and the Contributor tier costs far less, $0.10 input and $0.20 output per 1M tokens on Meta's list against $1.25 and $4.25. Meta's docs also keep the extended max reasoning level for the standard tier.

Install the official openai package, set base_url to https://api.zurelay.com/v1 and use your Zurelay API key. Then send a Chat Completions request with model set to "muse-spark-1.3-contributor"; Zurelay also accepts meta/muse-spark-1.3-contributor and muse-spark-1-3-contributor. Images go in image_url parts and PDFs in file parts.

1,048,576 tokens. On Zurelay it reads text, images and PDFs and writes text. Meta lists audio and video input for Muse Spark too, but Zurelay doesn't take them for this model: a request that includes one is turned away with a clear error before anything is billed.

The model always reasons before it answers. Set reasoning_effort to minimal, low, medium, high or xhigh; medium is the default. none isn't offered, and a request for none runs at minimal. Meta keeps the extended max level for standard Muse Spark 1.3, so a request for max runs at xhigh on this tier. In our tests it thinks a lot, and its thinking counts toward max_tokens: if the limit runs out while it is still thinking, the answer comes back empty, so leave room, a few thousand tokens for hard questions.

Yes. It calls tools you define, and Meta's API only takes tool_choice "auto". So when you send "required" or a named function, Zurelay sends auto, offers only the tool you named and tells the model to call it; for "none", it tells the model to answer in text.

It returns JSON when response_format asks for json_object or json_schema. Stop sequences have no effect: Meta's API ignores stop, so the answer runs to its natural end, and you can cut it in your own code if you need to.

Yes. Requests go to Meta's muse-spark-1.3-contributor, which Meta runs as the same model as muse-spark-1.3, at the Contributor tier. Zurelay doesn't swap in a different model. Only smart routing, which you turn on yourself, can answer with a close alternative when the model is down, and the x-zurelay-model header names the model that answered. Responses are sampled, so wording varies between runs. Zurelay is independent and is not affiliated with or endorsed by Meta.