Across 8 shared benchmarks, Gemma 4 E2B scores higher on 2 and InternVL-3-8B on 6. The widest gap is mathvision, where Gemma 4 E2B scores 52.4 against 29.6.
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.