Verdict
Built from the measurements below
- Accuracy is too close to call: Claude Sonnet 5.5 scored 78.8% and Muse Wonder 78.8% on the same 80 questions, a gap within what chance alone could produce.
- Claude Sonnet 5.5 is 3.9× cheaper per question as billed ($0.0054 vs $0.021), reasoning tokens included.
- Muse Wonder generates text 1.4× faster: 175 tokens per second against 126.
- Claude Sonnet 5.5 starts answering sooner: 0.7 s to the first token, against 16.1 s for Muse Wonder.
Which should you choose?
Choose Claude Sonnet 5.5 for the lower cost per correct answer ($0.0069 against $0.027), a quicker first token, shorter outputs (about 456 tokens per question against 1,014) and more hosts to choose from on OpenRouter (5 against 1). Choose Muse Wonder for faster streaming. The widest subject gap is in medicine: Claude Sonnet 5.5 answered 7 of 11 correctly, Muse Wonder 5 of 11. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.
Accuracy gap +0.0 points (z = 0.00). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.