Meta’s beta release of Muse Spark 1.2 includes a sharply discounted “contributor” version that allows data collection for product improvement. The pricing has drawn attention because the model is positioned as a code-focused upgrade and reportedly approaches leading systems on several coding benchmarks.
A steep price difference
The muse-spark-1.2-contributor model has a 1M-token context window and is priced at:
- $0.10 per million input tokens
- $0.002 per million cached input tokens
- $0.20 per million output tokens
The model is labeled “Used to improve our products.”
The standard muse-spark-1.2 version has the same 1M-token context window, but costs considerably more:
- $1.25 per million input tokens
- $0.15 per million cached input tokens
- $4.25 per million output tokens
That version is labeled “Not used to improve our products.”
The arrangement effectively gives developers a choice between lower inference costs and opting into data collection. Several commenters described the contributor pricing as attractive for experimentation and open-source work, while others questioned how the endpoint can be enabled in Meta’s beta Muse Code IDE.
Source: X