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

Liquid AI · released Aug 11, 2026

Wins 2 of 2 areas

Images and charts · Following instructions

Liquid AI · released Oct 22, 2025

Wins 0 of 2 areas

—

LFM2.5-VL-3B is the stronger all-rounder.

Scores updated Sep 10, 2026 · 32 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 video60LFM2.5-VL-3B6 of 6 tests
Following instructionsDoing exactly what it is asked10LFM2.5-VL-3B1 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 LFM2.5-VL-3B pulls ahead

Visual perception tasks people solve at a glanceBLINK+11.3points aheadLFM2.5-VL-3B61.5LFM2-VL-3B50.2
Math problems shown in pictures and chartsMathVista+6.4points aheadLFM2.5-VL-3B68.5LFM2-VL-3B62.1
Questions that truly need the image to answerMMStar+5.6points aheadLFM2.5-VL-3B63.3LFM2-VL-3B57.7

Where LFM2-VL-3B pulls ahead

No clear win on a test scored out of 100.

Every test, side by side

All 32 tests both models report. The winning score is in its model's colour; marks a score checked independently.

Images and chartsLFM2.5-VL-3B

BLINKLFM2.5-VL-3B by 11.361.550.2+11.3
MathVistaLFM2.5-VL-3B by 6.468.562.1+6.4
MMStarLFM2.5-VL-3B by 5.663.357.7+5.6
OCRBenchv2LFM2.5-VL-3B by 3.647.543.9+3.6
OCRBenchLFM2.5-VL-3B by 284.282.2+2
MMMU-ProLFM2.5-VL-3B by 1.830.528.7+1.8

Full images and charts ranking

Following instructionsLFM2.5-VL-3B

IFBenchLFM2.5-VL-3B by 525.820.8+5

Full following instructions ranking

Other results25 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.

ScreenSpot-v2 WebLFM2.5-VL-3B by 79.782.22.5+79.7
ScreenSpot-v2 MobileLFM2.5-VL-3B by 73.681.27.6+73.6
ScreenSpot-v2 DesktopLFM2.5-VL-3B by 72.778.76+72.7
RefCOCO (Macro Prec@1)LFM2.5-VL-3B by 30.887.957.1+30.8
BFCLv4LFM2.5-VL-3B by 1232.520.5+12
IFEvalLFM2.5-VL-3B by 9.482.372.9+9.4
MM-IFEvalLFM2.5-VL-3B by 9.260.651.4+9.2
OCRBench v2_enLFM2.5-VL-3B by 3.647.543.9+3.6
Multilingual MMBenchLFM2.5-VL-3B by 3.279.576.3+3.2
MMMU (val) (Pass@1)LFM2.5-VL-3B by 2.848.445.6+2.8
OCRBench v1LFM2.5-VL-3B by 2.584.281.7+2.5
InfographicVQA (val)LFM2.5-VL-3B by 2.470.267.8+2.4
SimpleVQALFM2.5-VL-3B by 2.435.433+2.4
RealWorldQALFM2.5-VL-3B by 273.171.1+2
DocVQA-valLFM2.5-VL-3B by 1.391.189.8+1.3
TextVQA-valLFM2.5-VL-3B by 1.384.383+1.3
MMMBLFM2.5-VL-3B by 1.18381.9+1.1
Multilingual MMMBLFM2.5-VL-3B by 1.18381.9+1.1
SEED-Bench (image)LFM2.5-VL-3B by 1.177.776.6+1.1
MMBenchLFM2.5-VL-3B by 18180+1
ChartQAtie81.380.4tie
ChartQA Testtie81.380.4tie
HallusionBenchtie47.246.4tie
POPEtie88.789.2tie
MMEtie73.173tie

Questions people ask

Which is better, LFM2.5-VL-3B or LFM2-VL-3B?

LFM2.5-VL-3B 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 32 benchmark tests both models have published scores on. The verdict counts the 7 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 25 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.