Across 6 shared benchmarks, Claude Opus 4.5 scores higher on 2 and Seed 2.0 on 4. The widest gap is browsecomp, where Seed 2.0 scores 77.3 against 37. Seed 2.0 is the cheaper of the two on tracked API pricing ($0.50 against $5.00 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 (anthropic-official, deepinfra), otherwise the lowest tracked offer.