Meta has introduced a low-cost Contributor tier for Muse Spark 1.2 and Muse Code. Let Meta use your prompts and model outputs to train future models, and you get a hefty discount.
At this tier, input, cache-hit, and output prices per million tokens are $0.10, $0.002, and $0.20, respectively. DeepSeek V4 Flash charges $0.14, $0.0028, and $0.28 — so Muse comes in roughly 29% cheaper across all three.
Don't want to share your data? Then you'll have to pay standard rates: $1.25 for input, $0.15 for cache hits, and $4.25 for output. That's 12.5x, 75x, and 21.25x the Contributor pricing, respectively.
The Contributor tier also has lower rate limits: max 60 requests per minute and 2.1 million tokens per minute. The standard tier, by contrast, allows 3,000 requests per minute and 4 million tokens per minute.
DeepSeek just teased a major API price increase, though exact numbers haven't been announced yet. Meta has already priced Muse 29% below DeepSeek's current rates — Zuckerberg is picking up the price-slasher baton.