Part of storyline · Meta Muse AI model releases
story 3 of 5Muse Spark Open adds a notable multimodal open-weights option with reported strengths on agent benchmarks, while licence conditions and uneven benchmark results constrain the breadth of the claim.
Limits of the evidence: Discovery-class source; benchmark figures and model specifications are reported without primary release documentation in the supplied text, although Artificial Analysis is identified as the benchmark source.
What to watch
Assessment revised Oct 23, 2026
Muse Spark Open leads U.S. open models: Reflection releases weights for Beam, a much-anticipated model
Andrew NgOct 19, 2026First report
First assessment, from Andrew Ng
No maturity ladder applies to this desk.
Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original sources are authoritative.