Across 45 shared benchmarks, Qwen3.5 27B scores higher on 39 and Qwen3 VL 235B A22B Reasoning on 6. The widest gap is HLE, where Qwen3.5 27B scores 48.5 against 13.6. Qwen3.5 27B is the cheaper of the two on tracked API pricing ($0.30 against $0.40 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.