Refresh Models —Deals —Regulatory —Sources —

InternVL-3-8B vs LFM2.5-VL-3B

9 SHARED BENCHMARKS

Across 9 shared benchmarks, InternVL-3-8B scores higher on 6 and LFM2.5-VL-3B on 3. The widest gap is MMMU (val) (Pass@1), where InternVL-3-8B scores 56.1 against 48.4.

OPENGVLABVSLIQUID AI9 SHARED6–3 HEAD-TO-HEAD

BenchmarkInternVL-3-8BMARGINLFM2.5-VL-3B

BLINK55.961.5

DocVQA-val9291.1

HallusionBench49.247.2

MathVista69.568.5

MMMU (val) (Pass@1)56.148.4

MMStar68.463.3

POPE90.688.7

RealWorldQA71.173.1

TextVQA-val82.284.3

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.