Verdict
Built from the measurements below
- Accuracy is too close to call: GPT-6.1 Sol scored 76.2% and GPT-6.1 Terra Pro 76.2% on the same 80 questions, a gap within what chance alone could produce.
- GPT-6.1 Sol is 1.7× cheaper per question as billed ($0.0029 vs $0.005), reasoning tokens included.
- GPT-6.1 Sol starts answering sooner: 2.1 s to the first token, against 8.5 s for GPT-6.1 Terra Pro.
Which should you choose?
Choose GPT-6.1 Sol for the lower cost per correct answer ($0.0037 against $0.0065), a quicker first token and shorter outputs (about 224 tokens per question against 453).
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.