LFM2.5-VL-3B：為邊緣運算帶來更快速、更卓越的視覺能力

Back to Articles LFM2.

5-VL-3B for Better and Faster Vision Capabilities for the Edge Team Article Published August 12, 2026 Upvote - Samuel Stevens samuelstevens Follow LiquidAI Ryan Shubert shubeydoo Follow LiquidAI Sina s-jse Follow LiquidAI Tianshu Yu tianshu-yu Follow LiquidAI Brandon Brandon3967 Follow LiquidAI Leonie Monigatti iamleonie Follow LiquidAI LFM2.

5-VL-3B is our most capable vision-language model you can run on your own hardware.It understands documents and screens alike, grounds objects, and can call tools.It answers directly instead of reasoning, so responses stay fast in real-time and on-device apps.LFM2.

5-VL-3B extends the vision-language capabilities of our previous releases with four major improvements: Screen/UI understanding: Strong understanding of digital screens across different devices.Grounding: Improved grounding and object detection with natural language queries.

Multi-image input: Improved reasoning across multiple images.Function calling: Significantly stronger at function calling, in text-only and vision-text situations.How we trained our most capable vision-language model LFM2.

5-VL-3B pairs a SigLIP2 400M NaFlex vision encoder with the same pre-trained backbone as our LFM2.5-2.6B text model.It is pre-trained on about 34T tokens, with 4x more vision data than before, drawn from curated and synthetic image-caption, OCR, grounding, and instruction-following sets.

To support non-Latin scripts, we doubled the vocabulary to 128K by extending the tokenizer in place rather than retraining from scratch.Post-training runs in two stages: First is supervised fine-tuning (SFT), with knowledge distillation from a larger teacher and Antidoom training.

Second is multi-reward reinforcement learning (RL).Benchmark results We evaluated LFM2.5-VL-3B across both vision and text benchmarks.

The vision benchmarks cover multilingual visual comprehension, instruction following, visual math and scientific reasoning, document understanding, object detection, multi-image understanding, and screen understanding.LFM2.

5-VL-3B leads its size class on real-world image tasks, while also reading digital content well, from documents and charts to on-screen UI elements.Task Benchmark LFM2.5-VL-3B (3.1B) LFM2-VL-3B (3.1B) gemma-4-E2B-it (5.1B) gemma-4-E4B-it (8B) InternVL 3.5 2B (2.4B) InternVL 3.5 4B (4.7B) Qwen3.

5-2B (2.3B) Qwen3.5-4B (4.7B) General MMStar 63.357.745.352.957.765.555.159.3 MME 73.173.054.967.673.681.076.279.5 RealWorldQA 73.171.160.064.361.667.765.167.1 SimpleVQA 35.433.027.330.430.533.735.240.7 SEED-Bench (image) 77.776.671.475.375.476.475.876.1 MMBench (dev EN v1.1) 81.080.064.271.676.281.

173.178.4 CountBenchQA 87.392.270.480.570.482.583.886.7 Multilingual MMMB 83.081.973.380.476.381.575.982.0 Multilingual MMBench 79.576.362.871.270.976.669.977.0 Multimodal IF MM-IFEval 60.651.465.668.247.154.555.463.1 STEM LogicVista 37.432.229.534.530.936.234.037.6 MathVista (mini) 68.562.137.845.

256.867.148.763.6 MMMU-Pro 30.528.726.932.621.322.724.936.0 MMMU (val) 48.445.641.149.352.060.744.150.3 Document, OCR & Chart ChartQA (test) 81.380.443.242.181.786.278.484.2 DocVQA (val) 91.189.885.787.488.491.892.694.8 InfographicVQA (val) 70.267.854.460.969.376.973.580.3 OCRBench v1 84.281.770.

273.583.982.084.485.6 OCRBench v2 (En) 47.543.944.448.845.549.147.758.7 TextVQA (val) 84.383.062.569.076.677.577.381.2 Grounding RefCOCO-avg 87.957.167.372.182.988.878.586.6 Multi-Image BLINK 61.550.245.252.252.057.248.658.7 MuirBench 58.334.932.951.845.053.548.262.0 Hallucination HallusionBench 47.

246.441.849.847.652.149.351.7 POPE 88.789.284.086.988.088.988.686.0 GUI ScreenSpot-v2 Desktop 78.76.028.145.879.982.063.876.3 ScreenSpot-v2 Mobile 81.27.642.960.386.287.869.781.4 ScreenSpot-v2 Web 82.22.522.447.679.982.665.977.8 Average - 69.457.252.059.764.669.463.770.

1 All values in the table are normalized to 0–100.Evaluation is done using vLLM 0.26.0 and each model’s recommended generation parameters when available.Non-reasoning mode is used everywhere, and models are prompted to directly answer without reasoning.We also evaluated LFM2.

5-VL-3B on text-only benchmarks for instruction following and tool use.Instruction following climbs across the board, and tool use improves sharply.On tool use, LFM2.5-VL-3B is on par with Gemma-4-E2B and Qwen3.5-2B.Task Benchmark LFM2.5-VL-3B (3.1B) LFM2-VL-3B (3.1B) gemma-4-E2B-it (5.

1B) gemma-4-E4B-it (8B) InternVL 3.5 2B (2.4B) InternVL 3.5 4B (4.7B) Qwen3.5-2B (2.3B) Qwen3.5-4B (4.7B) Instruction following IFEval 82.372.983.087.932.435.473.686.2 IFBench 25.820.834.139.224.424.528.933.5 Multi-IF 59.446.569.477.416.316.953.566.7 Tool use & function calling ToolSandbox 59.526.

456.561.6N/AN/A47.765.0 BFCL V4 32.520.533.240.0N/AN/A33.953.6 InternVL 3.5 models do not support function-calling.These results demonstrate that LFM2.5-VL-3B is a strong, general-purpose vision-language model.

It covers everyday tasks (captioning, visual question answering, document understanding) and is especially good at grounding objects, reading screens and documents, and calling tools.Inference speed on CPU and GPU LFM2.

5-VL-3B ships with day-one support across the inference ecosystem, including llama.cpp, MLX, vLLM, SGLang, and ONNX.On-device inference.LFM2.5-VL-3B decodes 228 tokens/s on an M5 Max and 116 tokens/s on a Ryzen AI Max+ 395, and fits in about 3 GB of memory.

It even reaches 20 tokens/s on a Galaxy S26 Ultra, so you can run it fully on-device.GPU inference.LFM2.5-VL-3B keeps latency consistently low and is the fastest on multi-frame inputs.LFM2.

5-VL-3B is also the fastest on output throughput out of all models we tested, reaching about 11K tokens per second at high concurrency.That is roughly 2× the larger 4B-class models and ahead of even the smaller 2B-class models, which adds up to nearly 1B output tokens per day on a single H100.

How to use LFM2.5-VL-3B Reach for LFM2.5-VL-3B when you need on-device intelligence for high-volume workloads.Install the latest version of transformers (compatible with transformers>=5.0.0): %pip install -q torch torchvision accelerate "transformers>=5.10.

1" Then load and run the model: import torch from transformers.imageutils import loadimage from transformers import AutoModelForImageTextToText, AutoProcessor from IPython.display import display MODELID = "LiquidAI/LFM2.5-VL-3B" processor = AutoProcessor.

frompretrained(MODELID) model = AutoModelForImageTextToText.frompretrained( MODELID, devicemap="auto", dtype="bfloat16", ) imgurl = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/cocosample.

png" inputimage = loadimage(imgurl) display(inputimage) messages = [ { "role": "user", "content": [ {"type": "image", "image": inputimage}, {"type": "text", "text": "Describe this image in two concise sentences."}, ], } ] inputs = processor.

applychattemplate( messages, addgenerationprompt=True, tokenize=True, returndict=True, returntensors="pt", ).to(model.device) with torch.inferencemode(): outputs = model.generate( **inputs, dosample=True, temperature=0.2, topk=50, repetitionpenalty=1.0, maxnewtokens=256, ) output = processor.

batchdecode(outputs[:, inputs["inputids"].shape[1]:], skipspecial_tokens=True)[0] print(output) Two cats are sleeping on a pink couch with two remote controls.You can find more hands-on examples on how to use LFM2.

5-VL3B for multi-image inputs, grounding, OCR, tool calling, and more in our documentation.Check out our release blog for video examples.LFM2.5-VL-3B demo Check out this browser demo of LFM2.5-VL-3B powering a vision-capable chat interface.

It allows you to take or upload multiple images and let the model interact with them, including grounding, OCR, and tool use.Get Started LFM2.5-VL-3B is available on Hugging Face today.With LFM2.5, we're delivering on our vision of AI that runs anywhere.These models are: Download: LFM2.

5-VL-3B on Hugging Face.Try: run the WebGPU demo in your browser, no setup needed.Fine-tune: adapt LFM2.5-VL-3B to your task with our fine-tuning tutorials.We can't wait to see what you build.Citation Please cite this article as: Liquid AI, "LFM2.

5-VL-3B: A Better and Faster Vision-Language Model for the Edge", Liquid AI Blog, Aug 2026.Or use the BibTeX citation: @article{liquidAI2026VL3B, author = {Liquid AI}, title = {LFM2.

5-VL-3B: A Better and Faster Vision-Language Model for the Edge}, journal = {Liquid AI Blog}, year = {2026}, note = {www.liquid.ai/blog/lfm2-5-vl-3b}, } Models mentioned in this article 2 More from this author Deploy local agents everywhere with LFM2.5-2.6B 87 August 4, 2026 LFM2.

5-Encoders for Fast Long-Context Inference on CPU 68 July 28, 2026 Community EditPreview Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.Tap or paste here to upload images Comment · Sign up or log in to comment Upvote - Models mentioned in this article 2