Across 24 shared benchmarks, Kimi K2 Instruct scores higher on 2 and Qwen3.7 Plus Preview on 22. The widest gap is HLE, where Qwen3.7 Plus Preview scores 35.6 against 7.9. Qwen3.7 Plus Preview is the cheaper of the two on tracked API pricing ($0.40 against $0.60 per million input tokens).
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (direct, alibaba-official), otherwise the lowest tracked offer.