Verdict
Built from the measurements below
- Accuracy is too close to call: Kimi K2.7 Code scored 65.0% and Step 5 Preview 63.8% on the same 80 questions, a gap within what chance alone could produce.
- Step 5 Preview is 1.3× cheaper per question as billed ($0.0047 vs $0.006), reasoning tokens included.
- Kimi K2.7 Code generates text 1.5× faster: 136 tokens per second against 93.
- Kimi K2.7 Code starts answering sooner: 0.4 s to the first token, against 0.6 s for Step 5 Preview.
- Step 5 Preview accepts more context: 1M tokens against 262k.
Which should you choose?
Choose Kimi K2.7 Code for faster streaming, a quicker first token and more hosts to choose from on OpenRouter (11 against 4). Choose Step 5 Preview for the lower cost per correct answer ($0.0074 against $0.0092) and a 1M-token context window.
Accuracy gap +1.3 points (z = 0.17). We call a winner only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.