Verdict
Built from the measurements below
- Accuracy is too close to call: DeepSeek V4.1 Flash 1017 scored 73.8% and Gemini 3.5 Flash Lite 75.0% on the same 80 questions, a gap within what chance alone could produce.
- DeepSeek V4.1 Flash 1017 is 3.7× cheaper per question as billed ($0.00091 vs $0.0034), reasoning tokens included.
- Gemini 3.5 Flash Lite generates text 7.1× faster: 409 tokens per second against 57.
- DeepSeek V4.1 Flash 1017 starts answering sooner: 0.9 s to the first token, against 1.1 s for Gemini 3.5 Flash Lite.
Which should you choose?
Choose DeepSeek V4.1 Flash 1017 for the lower cost per correct answer ($0.0012 against $0.0045), a quicker first token, open weights you can run yourself and more hosts to choose from on OpenRouter (30 against 2). Choose Gemini 3.5 Flash Lite for faster streaming. The widest subject gap is in medicine: Gemini 3.5 Flash Lite answered 7 of 11 correctly, DeepSeek V4.1 Flash 1017 4 of 11. 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 = -0.18). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.