VECTOR WIREAI INTELLIGENCE
UTC
Refresh Models Deals Regulatory Sources

InternVL3.5-8B-Instruct vs Qwen3.5 4B

13 SHARED BENCHMARKS

Across 13 shared benchmarks, InternVL3.5-8B-Instruct scores higher on 1 and Qwen3.5 4B on 12. The widest gap is Ego3D RMSE ↓, where InternVL3.5-8B-Instruct scores 23 against 13.2.

OPENGVLABVSALIBABA13 SHARED112 HEAD-TO-HEAD
CharXiv41.765.1
CountQA20.935.9
Ego3D RMSE ↓2313.2
ERQA4246.3
MMBench8087.1
MMMU6273.4
MMMU-Pro std46.464.9
MMMU-Pro vis42.361.3
MMStar64.175.3
OCRBench83.286.9
RealWorldQA66.976.3
SimpleVQA40.847.8
WaymoQA all58.167.1

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.