Which version is which?
| Version | Standard model ID | Muse Code access |
|---|---|---|
| 1.3 | muse-spark-1.3 | Pay-as-you-go and subscriptions |
| 1.2 | muse-spark-1.2 | Pay-as-you-go and subscriptions |
| 1.1 | muse-spark-1.1 | Not listed for Code |
Meta recommends Spark 1.3 for new work in its model catalog; 1.2 and 1.1 remain separately documented versions. The catalog describes a 1,048,576-token context window for these hosted models. Model availability for a particular team should still be checked in the actual account.
Standard and Contributor mode
The current pricing table lists the same Standard token rates for the three hosted versions: $1.25 input, $0.15 cached input and $4.25 output per million tokens. Contributor mode is listed for 1.3 and 1.2 with $0.10 input, $0.002 cached input and $0.20 output. Spark 1.1 has Standard mode only.
Contributor mode allows submitted content to be used for model training under the provider’s terms. It is a content-use choice as well as a billing choice. Do not switch a confidential workload simply because the tariff is lower. Search charges, when used, are additional.
These Spark rates do not establish Wonder Full rates or a Spark Open hosting quote. See Wonder versus Spark costs and self-hosted release costs for those distinct comparisons.
A migration that preserves a fair comparison
- Record the current version, mode, prompt, tools and reasoning setting.
- Choose a representative frozen task set with an explicit pass/fail rubric.
- Change the version first while keeping other settings fixed.
- Review errors, output quality, tool compatibility and actual usage.
- Test a mode change separately if your content policy permits it.
- Roll out gradually with a clear rollback identifier.
A higher version number does not guarantee a lower bill for your application. Different output lengths, retries and task success affect the total. Measure accepted work with consistent constraints, then compare actual spend.
Sources & evidence
Editorial check: 17 October 2026. Product facts come from the following primary sources; planning checklists and examples are our explanations. No paid inference, model benchmark or regional checkout was run for this guide.