Liquid AI model product
LFM3 VL 2B is an open-weight vision-language model that accepts text, images, and video and produces text.
Updated Nov 22, 2026. Default version: LFM3-VL-2B
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Benchmark scores for LFM3-VL-2B.
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkActivityNet | Score71.20% | Rank01 | Participants2 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkChartQA | Score92.70% | Rank01 | Participants27 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkDocVQA | Score96.70% | Rank01 | Participants29 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkInfoVQA | Score86.20% | Rank01 | Participants11 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMME | Score86.90% | Rank01 | Participants5 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMM IF-Eval | Score72.80% | Rank01 | Participants3 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkTextVQA | Score87.90% | Rank01 | Participants17 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkHR-Bench (4k) | Score82.30% | Rank02 | Participants2 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkPOPE | Score88.70% | Rank02 | Participants4 | Percentile66.67% | EvidenceC | Evaluated |
| BenchmarkRoboSpatialHome | Score0.63 points | Rank02 | Participants2 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkWe-Math | Score68.60% | Rank02 | Participants2 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkBLINK | Score71.10% | Rank05 | Participants18 | Percentile76.47% | EvidenceC | Evaluated |
| BenchmarkDynaMath | Score79.20% | Rank06 | Participants10 | Percentile44.44% | EvidenceC | Evaluated |
| BenchmarkMotionBench | Score64.00% | Rank06 | Participants6 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkRefCOCO-avg | Score0.91 points | Rank07 | Participants10 | Percentile33.33% | EvidenceC | Evaluated |
| BenchmarkEmbSpatialBench | Score0.83 points | Rank08 | Participants9 | Percentile12.50% | EvidenceC | Evaluated |
| BenchmarkHallusion Bench | Score65.60% | Rank08 | Participants21 | Percentile65.00% | EvidenceC | Evaluated |
| BenchmarkMMVU | Score52.90% | Rank08 | Participants8 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkCountBench | Score0.93 points | Rank09 | Participants10 | Percentile11.11% | EvidenceC | Evaluated |
| BenchmarkMLVU-M | Score71.30% | Rank09 | Participants9 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkOCRBench-V2 (en) | Score63.50% | Rank09 | Participants16 | Percentile46.67% | EvidenceC | Evaluated |
| BenchmarkRefSpatialBench | Score0.50 points | Rank09 | Participants9 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkCharXiv-D | Score87.50% | Rank10 | Participants19 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkMMMU (val) | Score55.70% | Rank11 | Participants14 | Percentile23.08% | EvidenceC | Evaluated |
| BenchmarkMuirBench | Score66.40% | Rank11 | Participants16 | Percentile33.33% | EvidenceC | Evaluated |
| BenchmarkVideoMME w/o sub. | Score62.80% | Rank11 | Participants11 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkMathVista-Mini | Score84.20% | Rank12 | Participants27 | Percentile57.69% | EvidenceC | Evaluated |
| BenchmarkMMBench-V1.1 | Score86.10% | Rank13 | Participants21 | Percentile40.00% | EvidenceC | Evaluated |
| BenchmarkVideoMME w sub. | Score69.00% | Rank13 | Participants13 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkMMStar | Score74.40% | Rank15 | Participants28 | Percentile48.15% | EvidenceC | Evaluated |
Preference and agent-evaluation results for the default version.
The default version does not have a matching Arena result yet.
Official vendor API pricing appears first, followed by individual provider offers.
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Provider-specific output speed and catalog latency for LFM3-VL-2B. Runtime does not affect the capability score.
No provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Technical details for the model's default version.
Available versions of this model. The score column identifies the version used in the overall ranking.
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionLFM3-VL-2B | Released | LLMBoard25.22 | Parameters1.9B | ContextN/A | Max outputN/A | Open weightsYes | LicenseLFM Open License v1.0 |
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
Key information about LFM3 VL 2B and its available data.
92B-parameter vision-language model. It processes text, images, and video sampled at 2 fps, with text output.
Data as of 2026-11-21.
Common questions about LFM3 VL 2B.
LFM3 VL 2B's default version was released on Oct 27, 2026.
No official standard PAYG price is currently available for LFM3 VL 2B.
LFM3 VL 2B was created by Liquid AI.
A context window is not available for the default version.
Yes. The default version is marked as open weight under LFM Open License v1.0.
No provider offering is currently linked to the default version.