Revision History

Recorded the September 4, 2026 public release of Muse Spark 1.3 max reasoning after the September 2 launch withheld it, the in-place edit of Meta's launch post, Meta's own 1.3 benchmark table with its methodology caveats, and independent Artificial Analysis measurements (pre- and post-Index v4.2) with dated snapshots.

Add Muse Spark 1.3 release, availability, efficiency, longer-horizon, and safety details

Added Meta's August 2026 multimodal evaluation, robotics demonstrations, WildArtifactBench preview, and independent evaluation context.

Correction: attribute Arena's Pareto-frontier remark to the model's placement, not its pricing

Add LMArena (Arena) Text Arena rank 4 result and independent WeirdML 60.3% result for Muse Spark 1.2; refresh Terminal-Bench absence check to August 8

Correction: DeepSWE 1.1 prose (GPT-5.6 Terra leads Muse Spark 1.2 there), three reference bylines, Wang 'rough edges' quote re-cited to his April 8 X thread, unsupported Bloomberg attribution removed, Avocado code name re-cited to CNBC.

Add Muse Spark 1.2 and Muse Code (released August 5, 2026): training approach, availability and pricing, Meta-reported benchmarks with harness caveats, and the Vals AI Finance Agent v2 result

Adds verified coverage of Muse Spark 1.1, including its agent and coding changes, 1 million-token context, public developer preview, benchmark results, pricing, and safety evaluation while preserving the existing article.

Added 6 contextual internal links

No summary provided