Across 15 shared benchmarks, Celeris-1 scores higher on 2 and MiniCPM5-2B on 12, with 1 level. The widest gap is TauBench V3 - Banking, where MiniCPM5-2B scores 20.8 against 3.9.
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.