README.md

metadata

library_name: transformers
license: other
license_name: lfm1.0
license_link: LICENSE
language:
  - ar
  - zh
  - en
  - fr
  - de
  - hi
  - id
  - it
  - ja
  - ko
  - pl
  - pt
  - ru
  - es
  - th
  - vi
pipeline_tag: image-text-to-text
base_model: LiquidAI/LFM2.5-VL-3B
tags:
  - liquid
  - lfm2.5
  - edge
  - decision
  - classification
  - calibration
  - system-one
  - multimodal

Try LFM • Docs • Discord

d1-3B

d1-3B is a 3B parameter decision model built on LFM2.5-VL-3B. You give it a state (text, JSON, images, or a mix) and a set of named questions. It returns calibrated, typed answers in one forward pass with zero output tokens: every answer is read directly from the model's distribution over the options, with no generation and no parsing.

• Best decision model under 10B: 47.12 on the Decision Index 0.2.1. That is ahead of every 4B and 9B model on the public leaderboard and level with Decider 35B-A3B (47.11).
• Multimodal: images and text in the same state. It scores 73.7 on 11 public image benchmarks (LFM2.5-VL-3B: 73.9).
• Fast: 8 ms a decision on an NVIDIA RTX 4090, 9 ms on an AMD MI325X, 30 ms on an Apple M5 Pro.

💻 Demos: Try d1-3B in a Hugging Face space without any setup: TODO: add demo

🗒️ Model Details

Model	Parameters	Description
LFM2.5-VL-3B	3.1B	General-purpose vision-language model (base)
d1-3B	3.1B	Post-trained for single-pass, calibrated decisions

d1-3B is a multimodal decision model with the following features:

• Total parameters: 3.12B
• LM backbone: LFM2.5-2.6B, 30 layers (22 double-gated short convolution blocks + 8 GQA)
• Vision encoder: SigLIP2 NaFlex shape-optimized 400M
• Context length: 32,768 tokens
• Vocabulary size: 128,000
• Output: no generated tokens. Each answer is a probability distribution over the question's options.

We recommend d1-3B wherever a pipeline needs a yes/no, a pick from named options, or a rating: routing and triage, moderation, intent and topic classification, extraction checks, reranking, LLM-as-a-judge scoring, agent guardrails, and visual inspection. It is not a chat model and does not write text.

🏃 How to use

Install the dependencies (requires transformers>=5.14):

pip install "transformers>=5.14" torch torchvision pillow

The model ships its own code, so load it with trust_remote_code=True:

import io
import urllib.request

import torch
from PIL import Image
from transformers import AutoModel

device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
model = AutoModel.from_pretrained("LiquidAI/d1-3b-RC", trust_remote_code=True,
                                  dtype=torch.float32 if device == "cpu" else torch.bfloat16).to(device)

# Several named questions over one text state, answered in one pass
questions = {
    "refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"},
    "team": {"type": "choice", "instructions": "Which team should handle this?",
             "criteria": {"billing": "Charges, refunds, invoices", "technical": "App or site faults",
                          "fraud": "Suspected unauthorised use"}},
    "urgency": {"type": "score", "instructions": "How urgent is this?",
                "criteria": ["Can wait", "Today", "Blocking the customer now"]},
}
print(model.system_one("I was charged twice this month, please refund one of them.", questions))

# An image as the whole state
url = "http://images.cocodataset.org/val2017/000000039769.jpg"  # two cats on a sofa
photo = Image.open(io.BytesIO(urllib.request.urlopen(url).read()))
print(model.system_one(None, {"cats": {"type": "choice", "instructions": "How many cats are there?",
                                       "criteria": {"one": "One", "two": "Two", "more": "Three or more"}}},
                       images=[photo]))

# Many requests, packed together with no padding
tickets = ["Where is my parcel? It was due Monday.", "The app crashes when I open settings."]
print(model.system_one_batch([(t, {"team": questions["team"]}) for t in tickets]))

call
system_one(state, questions, images=None)	Named questions over one state, in one pass. The state and its images are read once for all questions.
system_one_batch([(state, questions[, images]), ...])	Many requests, packed with no padding.

A state is a string, any JSON value, or None when the images are the whole state.

Questions and answers

Questions follow the Decision Index schema: type, instructions, and criteria.

type	criteria	answer fields
noul: yes or no	optional: {"true": "...", "false": "..."} to define each side	noul: P(yes)
choice: one of named options	{name: description}	choice, confidence, probabilities
score: 2 to 10 ordered levels	a list of level descriptions, lowest first	score (the expected level), confidence, probabilities, legend

Each call returns {"answers": {name: answer}, "usage": {"input_tokens": n, "output_tokens": 0}}.

⚡ Speed

Warm calls in bf16, median of 20 runs:

	one question	3 questions, one pass	3.4k-token state	384 px image	64 states, packed
NVIDIA RTX 4090	8.0 ms	21 ms	102 ms	17 ms	475 / s
AMD MI325X	9.2 ms	14 ms	44 ms	18 ms	1,106 / s
Apple M5 Pro (mps)	30 ms	41 ms	0.64 s	62 ms	78 / s

On NVIDIA GPUs, model.compile(mode="reduce-overhead") runs single questions as CUDA graphs. The RTX 4090 row uses it; without it a single question takes 16 ms. The first call with a new shape pays for kernel selection or compilation, so warm up the shapes you serve.

📊 Performance

All results are on public benchmarks.

Decision Index 0.2.1

The Decision Index casts 38 public benchmarks as 120,226 typed decisions, scored as chance-corrected skill. We scored d1-3B with the official scorer; it is not a leaderboard submission. All other rows come from the public leaderboard (snapshot 2026-09-28). Sizes are served parameters.

Model	Size	Decision Index	Knowledge	Language	Retrieval	Tools	Arts
Winnow-12B	12B	50.02	33.8	56.0	54.0	71.0	30.0
d1-3B	3B	47.12	27.8	53.5	52.2	66.7	34.6
Decider 35B-A3B	36B	47.11	31.8	55.5	54.7	56.5	32.6
JPT-9B	9.7B	46.89	31.7	56.7	44.6	67.0	28.6
Decision 1.0 Lux	9.7B	43.49	30.9	48.0	50.0	57.2	26.4
JPT-4B	4.7B	43.04	28.7	52.5	45.0	57.2	25.8
Jet v6.2	4.7B	42.60	28.7	43.9	48.2	62.9	27.0
Decider 4B	4.7B	40.70	25.7	46.0	44.7	58.6	25.0
Winnow-E4B	8.0B	39.89	22.3	45.1	43.8	62.5	22.8
Decider 2B	2.3B	28.97	14.9	32.6	37.3	42.4	14.6

Benchmarks as decisions

Evaluation splits are cast as typed decisions, with the same rows for every model. These scores follow the decision protocol, not each benchmark's own.

Benchmark	d1-3B	Decider 4B	Decider 2B
SQuAD 2.0	83.3	76.0	67.7
Civil Comments	93.3	92.8	93.6
MASSIVE intent	86.9	88.3	81.1
HelpSteer2	38.0	42.0	32.0
PubMedQA	68.3	63.3	65.7
BoolQ	86.3	89.0	87.3
XNLI	85.6	88.6	85.0
PAWS-X	76.4	69.8	59.5
Mean	77.3	76.2	71.5

d1-3B also scores 69.5 on DecisionBench (eng v1, all 23,900 rows) and 68.8 on Fast Decisions (dev split).

Vision

Eleven public image benchmarks, read as decisions over each benchmark's options (at most 1,000 rows each), compared with the base model:

Benchmark	d1-3B	LFM2.5-VL-3B
AI2D	80.1	80.9
BLINK	58.2	58.7
CV-Bench	82.7	87.6
HallusionBench	63.8	65.0
MMBench	84.3	84.3
MME	81.7	82.4
MMStar	59.4	61.2
MMVP	76.3	73.7
POPE	88.5	90.1
VisualWebBench	71.6	78.3
VL-RewardBench	63.6	50.9
Mean	73.7	73.9
ImajevBench (dev and calibration, 253 rows)	66.8	66.8

With the images removed, the same questions score 45.1, so the answers come from the images.

⚠️ Limitations

• Knowledge-heavy questions: larger models stay ahead on hard knowledge (for example MMLU-Pro and GPQA).
• Thresholds: calibrate any threshold on data from your own deployment before gating on it.
• No text output: d1-3B answers questions over fixed options. Use LFM2.5-VL-3B for open-ended generation.

📬 Contact

• Got questions or want to connect? Join our Discord community
• If you are interested in custom solutions with edge deployment, please contact our sales team.

Citation

@article{liquidai2025lfm2,
  title   = {LFM2 Technical Report},
  author  = {Liquid AI},
  journal = {arXiv preprint arXiv:2511.23404},
  year    = {2025}
}