Muse Spark

Released 2026-04 · reasoning · Not published · 1 benchmarks

Editorial notes

Lanzado 8 abril 2026. Primer modelo del nuevo Meta Superintelligence Labs (MSL). Closed/proprietary — ruptura con tradicion open-source de Meta. Modos Instant / Thinking / Contemplating. Reemplaza la apuesta frontier de Meta tras la pausa de Llama 4 Behemoth. Disponible en Meta AI app/web + private API preview a socios.

Spec sheet

Company: Meta
Country: US
Type: reasoning
Release: 2026-04
Context: Not published
Slug: muse-spark

Benchmarks (1)

Reasoning 1

Humanitys-Last-Exam

The hardest known benchmark — novel academic problems.

58.0

Cite this model

BibTeX · APA

BibTeX

@misc{frontier-muse-spark,
  title  = {Muse Spark},
  author = {{Meta}},
  year   = {2026},
  note   = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
  url    = {https://frontierbenchmarks.com/models/muse-spark}
}

APA

Meta (2026). Muse Spark [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/muse-spark

Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.

⚔️ Battle vs another model ← All models More from Meta Methodology