Compare / Claude Sonnet 5 vs Muse Spark 1.3

Claude Sonnet 5 vs Muse Spark 1.3

22 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.

Claude Sonnet 5
Anthropic · profile
Muse Spark 1.3
Meta · profile
Results compared
22
Shared sources
3
BenchmarkMetricClaude Sonnet 5Muse Spark 1.3
APEX-AgentsreportedTasks passed% of tasks 54.5%57.8%
Arena (formerly LMArena)reportedConfirmed task successIPS effect estimate 0.010.08
Arena (formerly LMArena)reportedPraise over complaintIPS effect estimate 0.090.04
Arena (formerly LMArena)reportedSteerabilityIPS effect estimate 0.070.02
Arena (formerly LMArena)reportedTool groundingIPS effect estimate 0.000.00
Arena (formerly LMArena)reportedOverallArena rating 1,4761,468
Arena (formerly LMArena)reportedOverallArena rating 1,4521,479
Arena (formerly LMArena)reportedBusiness, management and financeArena rating 1,4651,506
Arena (formerly LMArena)reportedCreative writingArena rating 1,4351,450
Arena (formerly LMArena)reportedExpert promptsArena rating 1,5121,527
Arena (formerly LMArena)reportedInstruction followingArena rating 1,4671,486
Arena (formerly LMArena)reportedOverallArena rating 1,4621,494
Arena (formerly LMArena)reportedWriting, literature and languageArena rating 1,4521,466
Artificial AnalysisreportedArtificial Analysis Coding Indexindex score 71.576.5
Artificial AnalysisreportedArtificial Analysis Intelligence Indexindex score 38.248.1
Artificial AnalysisreportedGPQA Diamond% of questions 91.1%94.1%
Artificial AnalysisreportedHumanity's Last Exam% of questions 41.3%48.7%
Artificial AnalysisreportedLong-context reasoning (AA-LCR)% of questions 82.0%83.0%
Artificial AnalysisreportedSciCode% of problems 54.3%59.7%
Artificial AnalysisreportedTerminal-Bench 2.1% of tasks 80.5%85.4%
Artificial AnalysisreportedTerminal-Bench 4.0% of tasks 14.1%33.3%
Artificial AnalysisreportedΤ-bench banking% of tasks 37.3%50.5%

Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.