lfm3-vl:latest

50 2 hours ago

LFM3-VL-2B is a vision-language model for on-device visual understanding.

vision tools thinking 2b 
ollama run lfm3-vl

Details

2 hours ago

280791ff33bb · 1.7GB

json
{ "architectures": [ "Lfm3VlForConditionalGeneration" ], "bos_token_id": 124894, "downsample_factor"
json
{ "_from_model_config": true, "bos_token_id": 124894, "do_sample": true, "eos_token_id": 124900, "ou
json
{ "producer": { "name": "modelopt", "version": "0.48.0rc1.dev167+g378888145" }, "quantization": { "q
json
{ "image_processor": { "do_convert_rgb": true, "do_normalize": true, "do_rescale": true, "do_resize"
json
{ "version": "1.0", "truncation": null, "padding": null, "added_tokens": [ { "id": 124893, "content"
json
{ "backend": "tokenizers", "bos_token": "<|startoftext|>", "eos_token": "<|im_end|>", "extra_special
license
LFM Open License v1.0 TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 1. Definitions. "
params
{ "repeat_last_n": -1, "repeat_penalty": 1.1, "temperature": 0.1, "top_k": 50, "
424 tensors

Readme

LFM3-VL-2B is a vision-language model for on-device visual understanding. It takes text, images, and video (soon) as input, and combines the LFM3 language model with a new in-house vision encoder.

New in LFM3-VL-2B:

  • Video understanding (soon): the first LFM vision model that takes video as input. Ask questions about a recorded lecture, a walkthrough filmed on your phone, or a clip from a camera feed.
  • Thinking: the model reasons step by step before it answers. This helps most with harder visual questions, such as math, charts, and multi-step problems.
  • Spatial and embodied reasoning: the model finds objects and reasons about how they are placed relative to each other, a strong foundation for robotics applications.

Model details

The LFM3 family includes base, text, vision, audio, omni, and decision models.

Property Specification
Checkpoint Post-trained vision-language model
Parameters 1.92B: 1.73B language model, 164M vision encoder, 25M projector
Architecture LFM3, 24 layers: 13 gated short convolutions, 5 sliding-window attention, 6 global attention
Vision encoder 14 layers: 8 gated depthwise 2D convolutions, 6 bidirectional attention, 16×16 patches with 2D rotary embeddings
Inputs Text, images, and video (soon) at native resolution. Each 2×2 block of patches becomes one language token. Video is sampled at 2 fps.
Context length 128K tokens (input + output)
Vocabulary 128,000 tokens
Languages English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, Polish
License LFM Open License v1.0

LFM3-VL-2B is designed for on-device visual tasks: questions about video, reading documents, charts, and app screens, finding objects in an image, and spatial reasoning for robots. It is not recommended for coding, multi-turn conversations, or complex reasoning.