LFM2.5-VL-3B vs Qwen3.5 4B

Liquid AI · released Aug 11, 2026

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.LFM2.5-VL-3B is better at images and charts; Qwen3.5 4B at following instructions.

Scores updated Sep 19, 2026 · 29 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 video32LFM2.5-VL-3B3 of 5 tests
Following instructionsDoing exactly what it is asked01Qwen3.5 4B1 of 1 test

Following instructions rests on a single test.

What it costs

Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.

LFM2.5-VL-3BNo current price is tracked.

Qwen3.5 4B$0.03 to read · $0.15 to write$0.18Price from Artificial Analysis

The biggest differences

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

Where LFM2.5-VL-3B pulls ahead

Math problems shown in pictures and chartsMathVista+4.9points aheadLFM2.5-VL-3B68.5Qwen3.5 4B63.6
Questions that truly need the image to answerMMStar+4points aheadLFM2.5-VL-3B63.3Qwen3.5 4B59.3
Visual perception tasks people solve at a glanceBLINK+2.8points aheadLFM2.5-VL-3B61.5Qwen3.5 4B58.7

Where Qwen3.5 4B pulls ahead

Harder college exam questions with imagesMMMU-Pro+34.9points aheadQwen3.5 4B65.4LFM2.5-VL-3B30.5
Follows unfamiliar, precisely checkable instructionsIFBench+26.2points aheadQwen3.5 4B52LFM2.5-VL-3B25.8
Reads text in imagesOCRBench+2.7points aheadQwen3.5 4B86.9LFM2.5-VL-3B84.2

Every test, side by side

All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.

Images and chartsLFM2.5-VL-3B

MMMU-ProQwen3.5 4B by 34.930.565.4+34.9
MathVistaLFM2.5-VL-3B by 4.968.563.6+4.9
MMStarLFM2.5-VL-3B by 463.359.3+4
BLINKLFM2.5-VL-3B by 2.861.558.7+2.8
OCRBenchQwen3.5 4B by 2.784.286.9+2.7

Full images and charts ranking

Following instructionsQwen3.5 4B

IFBenchQwen3.5 4B by 26.225.852+26.2

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.

BFCLv4Qwen3.5 4B by 18.132.550.6+18.1
OCRBench v2_enQwen3.5 4B by 11.247.558.7+11.2
InfographicVQA (val)Qwen3.5 4B by 10.170.280.3+10.1
MMEQwen3.5 4B by 6.473.179.5+6.4
RealWorldQALFM2.5-VL-3B by 673.167.1+6
SimpleVQAQwen3.5 4B by 5.335.440.7+5.3
HallusionBenchQwen3.5 4B by 4.547.251.7+4.5
ScreenSpot-v2 WebLFM2.5-VL-3B by 4.482.277.8+4.4
IFEvalQwen3.5 4B by 3.982.386.2+3.9
DocVQA-valQwen3.5 4B by 3.791.194.8+3.7
TextVQA-valLFM2.5-VL-3B by 3.184.381.2+3.1
ChartQA TestQwen3.5 4B by 2.981.384.2+2.9
POPELFM2.5-VL-3B by 2.788.786+2.7
MMBenchLFM2.5-VL-3B by 2.68178.4+2.6
MM-IFEvalQwen3.5 4B by 2.560.663.1+2.5
Multilingual MMBenchLFM2.5-VL-3B by 2.579.577+2.5
ScreenSpot-v2 DesktopLFM2.5-VL-3B by 2.478.776.3+2.4
MMMU (val) (Pass@1)Qwen3.5 4B by 1.948.450.3+1.9
SEED-Bench (image)LFM2.5-VL-3B by 1.677.776.1+1.6
OCRBench v1Qwen3.5 4B by 1.484.285.6+1.4
MMMBLFM2.5-VL-3B by 18382+1
Multilingual MMMBLFM2.5-VL-3B by 18382+1
ScreenSpot-v2 Mobiletie81.281.4tie

Questions people ask

Which is better, LFM2.5-VL-3B or Qwen3.5 4B?

LFM2.5-VL-3B and Qwen3.5 4B each win one of the two areas where both have results. LFM2.5-VL-3B wins images and charts; Qwen3.5 4B wins following instructions.

How do you compare the two?

We use the 29 benchmark tests both models have published scores on. The verdict counts the 6 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.