Across 37 shared benchmarks, MiniMax M2.7 scores higher on 18 and Qwen3.5 27B on 19. The widest gap is OmniScience Non-Hallucination, where MiniMax M2.7 scores 65.6 against 24.9. Both cost $0.30 per million input tokens on tracked API pricing; MiniMax M2.7 is cheaper on output ($1.20 against $2.40).
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, direct), otherwise the lowest tracked offer.