LFM3-VL-2B is the stronger all-rounder.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the one area 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.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where LFM3-VL-2B pulls ahead
Where LFM2.5-VL-3B pulls ahead
No clear win on a test scored out of 100.
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 chartsLFM3-VL-2B
Other results24 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.
- Math & logical reasoningLFM3-VL-2B by 28.240.668.8+28.2
- HallusionBenchLFM3-VL-2B by 18.447.265.6+18.4
- InfographicVQA (val)LFM3-VL-2B by 1670.286.2+16
- OCRBench v2_enLFM3-VL-2B by 1647.563.5+16
- ScreenSpot-v2 DesktopLFM3-VL-2B by 15.978.794.6+15.9
- MMELFM3-VL-2B by 13.873.186.9+13.8
- Document, OCR & ChartLFM3-VL-2B by 12.771.283.9+12.7
- MM-IFEvalLFM3-VL-2B by 12.260.672.8+12.2
- ChartQA TestLFM3-VL-2B by 11.481.392.7+11.4
- ScreenSpot-v2 MobileLFM3-VL-2B by 1181.292.2+11
- GeneralLFM3-VL-2B by 8.457.666+8.4
- PerceptionLFM3-VL-2B by 7.767.975.6+7.7
- ScreenSpot-v2 WebLFM3-VL-2B by 7.782.289.9+7.7
- MMMU (val) (Pass@1)LFM3-VL-2B by 7.348.455.7+7.3
- RealWorldQALFM3-VL-2B by 6.273.179.3+6.2
- DocVQA-valLFM3-VL-2B by 5.691.196.7+5.6
- MMBenchLFM3-VL-2B by 5.18186.1+5.1
- Multilingual MMBenchLFM3-VL-2B by 4.179.583.6+4.1
- TextVQA-valLFM3-VL-2B by 3.684.387.9+3.6
- MMMBLFM3-VL-2B by 2.28385.2+2.2
- OCRBench v1LFM3-VL-2B by 2.184.286.3+2.1
- SEED-Bench (image)tie77.778.6tie
- SimpleVQAtie35.436.3tie
- POPEtie88.788.7tie
Questions people ask
Which is better, LFM2.5-VL-3B or LFM3-VL-2B?
LFM3-VL-2B wins the one area where both have results: images and charts. LFM2.5-VL-3B wins none.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 4 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.