Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

InternVL3_5-4B vs LFM2-VL-3B

OpenGVLab

Wins 2 of 2 areas

Images and charts · Following instructions

Liquid AI · released

Wins 0 of 2 areas

—

InternVL3_5-4B is the stronger all-rounder.

Scores updated · 27 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.

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

  • Questions that truly need the image to answerMMStar+7.8points aheadInternVL3_5-4B65.5LFM2-VL-3B57.7
  • Visual perception tasks people solve at a glanceBLINK+7points aheadInternVL3_5-4B57.2LFM2-VL-3B50.2
  • Math problems shown in pictures and chartsMathVista+5points aheadInternVL3_5-4B67.1LFM2-VL-3B62.1

Where LFM2-VL-3B pulls ahead

  • Harder college exam questions with imagesMMMU-Pro+6points aheadLFM2-VL-3B28.7InternVL3_5-4B22.7

Every test, side by side

All 27 tests both models report. The winning score is in its model's colour; marks a score checked independently.

Images and chartsInternVL3_5-4B
  • MMStarInternVL3_5-4B by 7.865.557.7+7.8
  • BLINKInternVL3_5-4B by 757.250.2+7
  • MMMU-ProLFM2-VL-3B by 622.728.7+6
  • MathVistaInternVL3_5-4B by 567.162.1+5

Full images and charts ranking

Following instructionsInternVL3_5-4B
  • IFBenchInternVL3_5-4B by 3.724.520.8+3.7

Full following instructions ranking

Other results22 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.

Questions people ask

Which is better, InternVL3_5-4B or LFM2-VL-3B?

InternVL3_5-4B wins all two areas where both have results: images and charts and following instructions. LFM2-VL-3B wins none.

How do you compare the two?

We use the 27 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 22 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.