Muse Spark 1.3
Muse Spark 1.3 improves agentic and coding work, with about 20% fewer tool calls and 25% fewer tokens, and it asks clarifying questions before consequential actions.
An efficiency and reliability release with no new capability class. It has better long-instruction adherence, self-knowledge of limits and adversarial resistance, and is available in Muse Code and Meta Model API. Zuckerberg again promised open weights 'soon'.
- Date
- Wednesday, 2 September 2026
- Lab
- Meta
- Kind
- model
- Access
- closed API
Figures
| Measure | Value | Measured by |
|---|---|---|
| Coding efficiency vs 1.2 | ~20% fewer tool calls, ~25% fewer tokens Meta-reported | company |
The Register cites Artificial Analysis placing 1.3 level with GPT 5.6 Sol, Claude Opus 5 and Grok 4.6 High (not independently opened).
Sources
- research.meta.ai/blog/introducing-muse-spark-1-3
- www.theregister.com/ai-and-ml/2026/09/02/zucks-muse-to-spark-joy-with-open-weights-release
This record was checked against its sources on 6 October 2026. How we check