Verdict
Built from the measurements below
- Accuracy is too close to call: Claude Sonnet 5.5 scored 78.8% and Gemini 3.1 Pro Preview 77.5% on the same 80 questions, a gap within what chance alone could produce.
- Claude Sonnet 5.5 is 3.2× cheaper per question as billed ($0.0054 vs $0.017), reasoning tokens included.
- Output speed is similar (126 vs 124 tokens per second).
- Claude Sonnet 5.5 starts answering sooner: 0.7 s to the first token, against 3.4 s for Gemini 3.1 Pro Preview.
Which should you choose?
Choose Claude Sonnet 5.5 for the lower cost per correct answer ($0.0069 against $0.022), a quicker first token, shorter outputs (about 456 tokens per question against 1,382) and more hosts to choose from on OpenRouter (5 against 2). The widest subject gap is in engineering: Claude Sonnet 5.5 answered 13 of 15 correctly, Gemini 3.1 Pro Preview 10 of 15. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.
Accuracy gap +1.2 points (z = 0.19). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.