InternVL3_5-4B vs Qwen3.5 2B

OpenGVLab

Wins 1 of 2 areas

Images and charts

Alibaba · released Mar 2, 2026

Wins 1 of 2 areas

Following instructions

The two are evenly matched.InternVL3_5-4B is better at images and charts; Qwen3.5 2B at following instructions.

Scores updated Sep 19, 2026 · 28 tests both models report · How we compare

Where each one wins

Tests won in each of the two areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.

Images and chartsUnderstanding pictures, charts and video31InternVL3_5-4B3 of 4 tests
Following instructionsDoing exactly what it is asked01Qwen3.5 2B1 of 1 test

Following instructions rests on a single test.

The biggest differences

The tests each model wins by the widest margin, up to three each. Scores are out of 100.

Where InternVL3_5-4B pulls ahead

Math problems shown in pictures and chartsMathVista+18.4points aheadInternVL3_5-4B67.1Qwen3.5 2B48.7
Questions that truly need the image to answerMMStar+10.4points aheadInternVL3_5-4B65.5Qwen3.5 2B55.1
Visual perception tasks people solve at a glanceBLINK+8.6points aheadInternVL3_5-4B57.2Qwen3.5 2B48.6

Where Qwen3.5 2B pulls ahead

Harder college exam questions with imagesMMMU-Pro+20.3points aheadQwen3.5 2B43InternVL3_5-4B22.7
Follows unfamiliar, precisely checkable instructionsIFBench+7points aheadQwen3.5 2B31.5InternVL3_5-4B24.5

Every test, side by side

All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.

Images and chartsInternVL3_5-4B

MMMU-ProQwen3.5 2B by 20.322.743+20.3
MathVistaInternVL3_5-4B by 18.467.148.7+18.4
MMStarInternVL3_5-4B by 10.465.555.1+10.4
BLINKInternVL3_5-4B by 8.657.248.6+8.6

Full images and charts ranking

Following instructionsQwen3.5 2B

IFBenchQwen3.5 2B by 724.531.5+7

Full following instructions ranking

Other results23 tests, not counted

Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.

IFEvalQwen3.5 2B by 38.235.473.6+38.2
ScreenSpot-v2 DesktopInternVL3_5-4B by 18.28263.8+18.2
ScreenSpot-v2 MobileInternVL3_5-4B by 18.187.869.7+18.1
ScreenSpot-v2 WebInternVL3_5-4B by 16.782.665.9+16.7
MMMU (val) (Pass@1)InternVL3_5-4B by 16.660.744.1+16.6
HallusionBenchQwen3.5 2B by 13.452.165.5+13.4
MMBenchInternVL3_5-4B by 881.173.1+8
ChartQA TestInternVL3_5-4B by 7.886.278.4+7.8
MMMBInternVL3_5-4B by 781.574.5+7
Multilingual MMBenchInternVL3_5-4B by 6.776.669.9+6.7
Multilingual MMMBInternVL3_5-4B by 5.681.575.9+5.6
MMEInternVL3_5-4B by 4.88176.2+4.8
InfographicVQA (val)InternVL3_5-4B by 3.476.973.5+3.4
RealWorldQAInternVL3_5-4B by 2.667.765.1+2.6
OCRBench v1Qwen3.5 2B by 2.48284.4+2.4
SimpleVQAQwen3.5 2B by 1.533.735.2+1.5
OCRBench v2_enInternVL3_5-4B by 149.148.1+1
MM-IFEvaltie54.555.4tie
DocVQA-valtie91.892.6tie
SEED-Bench (image)tie76.475.8tie
VLM Judge Scoretie90.289.7tie
POPEtie88.988.6tie
TextVQA-valtie77.577.3tie

Questions people ask

Which is better, InternVL3_5-4B or Qwen3.5 2B?

InternVL3_5-4B and Qwen3.5 2B each win one of the two areas where both have results. InternVL3_5-4B wins images and charts; Qwen3.5 2B wins following instructions.

How do you compare the two?

We use the 28 benchmark tests both models have published scores on. The verdict counts the 5 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 23 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.