VECTOR WIREAI INTELLIGENCE
⌘K
UTC
Refresh Models —Deals —Regulatory —Sources —

InternVL3.5-8B-Instruct vs Qwen Drive 1.0 SFT

13 SHARED BENCHMARKS

Across 13 shared benchmarks, InternVL3.5-8B-Instruct scores higher on 1 and Qwen Drive 1.0 SFT on 12. The widest gap is Ego3D RMSE ↓, where InternVL3.5-8B-Instruct scores 23 against 7.8.

OPENGVLABVSALIBABA13 SHARED1–12 HEAD-TO-HEAD
BenchmarkInternVL3.5-8B-InstructMARGINQwen Drive 1.0 SFT
CharXiv41.764.4
CountQA20.931.7
Ego3D RMSE ↓237.8
ERQA4248.5
MMBench8085.5
MMMU6272.7
MMMU-Pro std46.462.7
MMMU-Pro vis42.359.7
MMStar64.175.9
OCRBench83.286.4
RealWorldQA66.979
SimpleVQA40.846.1
WaymoQA all58.174.5

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.