Benchmark Almanac

LFM2.5-VL-3B

LFM2.5-VL-3B places in the top three on POPE.

Liquid AI Released August 12th, 2026 Family LFM2.5 The lab's model card Compare with others
29
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Liquid AI Aug 2026

LFM2.5-VL-3B

Open weights Small

120.8 #340 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • MMMU-Pro 30.5% #88 /90
  • IFBench 25.8% #77 /80
  • BFCL V4 32.5% #29 /30

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
POPE 88.7% #2 of 7 Lab model card
MMMB 83% #1 of 2 Lab model card
MM-IFEval 60.6% #2 of 3 Lab model card
IFEval 82.3% #97 of 135 Lab model card
Multi-IF 59.4% #23 of 28 Lab model card
IFBench 25.8% #77 of 80 Lab model card
SimpleVQA 35.4% #29 of 30 Lab model card
HallusionBench 47.2% #30 of 31 Lab model card

Agents and tools

BenchmarkScoreRankSource
ScreenSpot-v2 80.7% #13 of 14 Lab model card
BFCL V4 32.5% #29 of 30 Lab model card
ToolSandbox 59.5% #5 of 5 Lab model card

Vision, audio and robotics

BenchmarkScoreRankSource
TextVQA 84.3% #4 of 17 Lab model card
SEED-Bench 77.7% #1 of 3 Lab model card
BLINK 61.5% #18 of 29 Lab model card
RefCOCO / RefCOCO+ / RefCOCOg 87.9% #20 of 29 Lab model card
RealWorldQA 73.1% #27 of 38 Lab model card
MathVista 68.5% #50 of 70 Lab model card
InfographicVQA 70.2% #19 of 26 Lab model card
DocVQA 91.1% #27 of 36 Lab model card
ChartQA 81.3% #29 of 37 Lab model card
CountBench 87.3% #22 of 28 Lab model card
MuirBench 58.3% #10 of 12 Lab model card
OCRBench 84.2% #29 of 34 Lab model card
MMBench-EN v1.1 81% #21 of 23 Lab model card
MMStar 63.3% #33 of 36 Lab model card
LogicVista 37.4% #13 of 14 Lab model card
MMMU 48.4% #98 of 103 Lab model card
MMMU-Pro 30.5% #88 of 90 Lab model card
OCRBench v2 42.6% #14 of 14 Official leaderboard

Its lab's numbers put it at 120.8, number 340 of 404

The Self-Reported Index fits one score to the 24 capability results Liquid AI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 118.1 to 123.6. How we compute it.

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
OCRBench v2 47.5% OCRBench v2 (En) 42.6% Leaderboard +4.9 pts

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

LFM2.5-VL-3B (Liquid AI blog) Aug 12, 2026