Verdict
Built from the measurements below
- Accuracy is too close to call: Granite 4.0 Micro scored 25.0% and Nova Micro 1.0 22.5% on the same 80 questions, a gap within what chance alone could produce.
- Granite 4.0 Micro is 1.7× cheaper per question as billed ($0.000051 vs $0.000087), reasoning tokens included.
- Nova Micro 1.0 generates text 4.7× faster: 180 tokens per second against 38.
- Nova Micro 1.0 starts answering sooner: 0.5 s to the first token, against 0.7 s for Granite 4.0 Micro.
Which should you choose?
Choose Granite 4.0 Micro for the lower cost per correct answer ($0.0002 against $0.00039) and open weights you can run yourself. Choose Nova Micro 1.0 for faster streaming and a quicker first token. The widest subject gap is in economics: Nova Micro 1.0 answered 6 of 9 correctly, Granite 4.0 Micro 2 of 9. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.
Accuracy gap +2.5 points (z = 0.37). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.