Pricing and budget · 2026-09-22
Muse Spark 1.3 Contributor: the price ratio in exchange for the data policy
Covers that meta/muse-spark-1.3-contributor carries the same capabilities as the standard 1.3 checkpoint, that you must grant permission for your prompts and completions to train Meta's models in exchange, and which jobs fit this tier.
The same capability set, a different condition
meta/muse-spark-1.3-contributor carries the same 1,048,576-token context window, the same image/video/PDF input support, the same function calling, JSON mode, and structured output capabilities, and the same minimal-to-xhigh reasoning_effort range as standard muse-spark-1.3. The difference isn't output quality; the catalog description says this directly: the contributor tier offers the same checkpoint at heavily discounted rates, in exchange for granting Meta permission to use your prompts and completions to train future models.
This is the same trade-off pattern seen in the Muse Spark 1.2 pair (the contributor tier there also cut input/output price by multiple times); the same logic holds for the 1.3 pair, only the checkpoint version differs.
Which jobs this tier makes sense for
The catalog page explicitly scopes this tier: it fits prototyping, load testing, and integration trials — work whose data isn't sensitive. Sending a job containing confidential, personal, or customer data to this tier means granting permission for that data to be used by Meta to train models; that kind of work should stay on the standard muse-spark-1.3 tier.
This tier's rate limit is also low and shared across all LLMTR users; that means it isn't a fit for high-volume production traffic, but works fine for a low-volume prototype or trial flow.
- Fits: prototyping, load testing, non-sensitive integration trials.
- Doesn't fit: work containing confidential, personal, or customer data.
- Rate limit is low and shared across the platform; not for high-volume production.
Decide based on data sensitivity, not price
As attractive as the discount rate is, choosing this tier is primarily a data-policy decision; whether your job's data is acceptable to use as Meta's training data should come before the price difference. If the answer is no, you need to use the standard tier no matter how large the discount is.
Frequently asked questions
Is Muse Spark 1.3 Contributor's output quality lower than standard?
No, it runs the same checkpoint and there's no output-quality difference; the only difference is price and the data usage policy.
Can I change my mind and withdraw my data after choosing this tier?
That depends on Meta's data usage policy; check the provider's current data policy documentation for exact terms, as LLMTR doesn't alter that policy.