[![](https://huggingface.co/LiquidAI/d1-2B/resolve/main/assets/icons/api.svg)API](https://console.liquid.ai/dashboard)
[![](https://huggingface.co/LiquidAI/d1-2B/resolve/main/assets/icons/try.svg)Try](https://playground.liquid.ai/)
[![](https://huggingface.co/LiquidAI/d1-2B/resolve/main/assets/icons/docs.svg)Docs](https://docs.liquid.ai/lfm/getting-started/welcome)
[![](https://huggingface.co/LiquidAI/d1-2B/resolve/main/assets/icons/blog.svg)Blog](https://www.liquid.ai/blog/open-d1)

d1-2B

d1-2B is a decision model for text, pictures, and video, built on LFM3-VL-2B. It returns calibrated answers in one forward pass, without generating tokens.

• Best-in-class: it leads the open d1 models on average and on our board of over a hundred decision benchmarks, ahead of d1-3B at under two thirds of its size.
• Beyond classification: it also extracts text, numbers, and dates with their span, locates objects in pictures, and estimates quantities with an interval.
• Millisecond decisions: 5.7 ms per decision on an NVIDIA H100 and 3.8 ms on an AMD MI325X, with any number of questions about one state in one call.

Find more information about open d1 in our release blog post.

Model details

The LFM3 family includes base, text, vision, audio, omni, and decision models. d1-2B decides over text, pictures, and video, and d1-Omni-2B adds audio.

Property Specification

Checkpoint Post-trained decision model, built on LFM3-VL-2B

Parameters 1.92B

Architecture LFM3 backbone, 24 layers: 13 gated short convolutions, 5 sliding-window attention, 6 global attention

Context length 128K tokens

Vocabulary 128,000 tokens

Input A state (text or JSON), pictures or a video, and any number of named questions

Output Typed answers with their probabilities, with no generated tokens

Languages English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, Polish

License LFM Open License v1.0

d1-2B is designed for any pipeline step that needs a decision: routing and triage, moderation and guardrails, intent and topic classification, extraction from documents and forms, judging and reranking, and visual inspection. It is not a chat model.

Knowledge-heavy questions favor larger models, and answers get less reliable past about 64K tokens of state. Estimates are reliable on counts and scores, less so on free quantities.

Inference

d1-2B runs in Transformers with its own model code. It reads its answers from a single forward pass, so there are no sampling parameters to set.

Transformers quick start:

pip install "transformers>=5.19" torch torchvision pillow

import torch
from transformers import AutoModel

device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
dtype = torch.float32 if device == "cpu" else torch.bfloat16
model = AutoModel.from_pretrained("LiquidAI/d1-2B", trust_remote_code=True, dtype=dtype).to(device)

# Several named questions over one state, answered in one pass
questions = {
    "refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"},  # yes/no: P(yes)
    "team": {"type": "choice", "instructions": "Which team should handle this?",
             "criteria": {"billing": "Charges, refunds, invoices", "shipping": "Delivery and parcels",
                          "technical": "App or site faults"}},
    "topics": {"type": "choice", "multiple": True, "instructions": "Which topics does the message raise?",
               "criteria": {"billing": "Payments and charges", "delivery": "Shipping and parcels",
                            "login": "Signing in to the account"}},
    "urgency": {"type": "score", "instructions": "How urgent is this ticket?", "criteria": ["low", "medium", "high"]},
}
out = model.system_one("I was charged twice and my parcel never arrived. Please refund one charge.", questions)
a = out["answers"]
urgency = a["urgency"]["legend"][str(int(a["urgency"]["score"]))]
print(round(a["refund"]["noul"], 2), a["team"]["choice"], a["topics"]["choices"], urgency)
# 1.0 billing ['billing', 'delivery'] medium

Values from a document

invoice = "ACME SA\nInvoice 2026-117\nTotal due: 1 234,50 EUR\nPay before 03/04/2026 (day/month/year)"
out = model.system_one(invoice, {
    "payee": {"type": "text", "instructions": "Who issued the invoice?"},
    "total": {"type": "number", "instructions": "How much is due?"},
    "due": {"type": "date", "instructions": "By when must it be paid?"},
    "vat": {"type": "number", "instructions": "How much VAT is charged?"},
})
a = out["answers"]
print(a["payee"]["text"], a["total"]["number"], a["due"]["date"], a["vat"]["number"])
# ACME SA 1234.5 2026-04-03 None

Pictures and boxes

Count, locate one object, or list every instance (max reads at most that many):

from transformers.image_utils import load_image

image = load_image("http://images.cocodataset.org/val2017/000000039769.jpg")  # two cats on a sofa
out = model.system_one(None, {
    "cats": {"type": "number", "instructions": "How many cats are there?", "range": [0, 10], "step": 1},
    "remote": {"type": "box", "instructions": "Where is the remote control on the left?"},
    "every_cat": {"type": "boxes", "instructions": "Every cat.", "max": 4},
}, media=[image])
a = out["answers"]
print(a["cats"]["number"], a["every_cat"]["count"], {k: round(v, 2) for k, v in a["remote"]["box"].items()})
# 2.0 2 {'top': 0.15, 'bottom': 0.24, 'left': 0.06, 'right': 0.28}

Many requests in one call

tickets = ["Where is my parcel? It was due Monday.", "The app crashes when I open settings."]
for out in model.system_one_batch([(t, {"team": questions["team"]}) for t in tickets]):
    print(out["answers"]["team"]["choice"])
# shipping
# technical

Calls

Call What it does

system_one(state, questions, media=None) Named questions over one state, in one pass. media lists pictures (PIL images or {"type": "image", "path": ...}) and videos as their frames ({"type": "video", "video": [PIL images], "fps": 24}).

system_one_batch([(state, questions[, media]), ...]) Many requests in one call, batched.

set_size= (either call) Questions read together after the state. The default is the model's trained size (16). Larger is faster and costs some accuracy on text.

A state is a string, any JSON value, or None when the media are the whole state. Each call returns {"answers": {name: answer}, "usage": {"input_tokens": n, "output_tokens": 0}}.

Question types

type Question Answer

noul Yes or no noul: P(yes)

choice One of the options criteria names ({name: description}) choice, confidence, probabilities

choice with multiple: true Any of the options criteria names choices, probabilities (each option's own P(yes))

score One of 2 to 10 ordered levels, with criteria listing them lowest first score (the median level), confidence, probabilities, legend

text, number, date A value to read from the state The field of its type (a number with its unit, a date in ISO 8601), span, found

number with range An estimate, step apart number (the median), interval (p10, p90)

box, boxes One object, or every instance (16 at most, or max), in the one picture box (top, bottom, left, right as fractions of the picture), found, confidence, or boxes and count

A value or a box is missing when found is under 0.5, and it is then None. Calibrate any decision threshold on data from your own use.

Performance

Benchmarks

d1-2B leads the open d1 models on our decision board, on average, and on most single benchmarks, ahead of d1-3B with under two thirds of its parameters.

Benchmark d1
2B d1-omni
600M d1-Omni
2B d1
3B

Decisions

Decision Index public suite (v0.3) 51.1 18.3 49.0 49.0

DecisionBench (Hanno Labs) v1 77.1 58.7 76.4 73.1

Decision Bench (decisionbench.ai) v4 85.2 57.9 84.7 83.6

Fast Decisions 58.8 – 56.8 59.3

JevBench v1.4.1 77.9 58.0 75.3 76.2

ImajevBench v2.0-lite 75.5 21.7 72.3 64.0

WorkflowEvals 81.6 48.8 77.1 74.4

Typed Decisions 61.3 57.5 63.1 65.0

Our board (over a hundred public and private benchmarks) 63.2 – 61.8 58.8

Vision

AI2D 76.7 37.1 71.5 79.9

BLINK 66.2 37.7 63.5 59.2

CountQA 51.4 26.8 45.4 42.6

CV-Bench 86.1 40.8 84.0 82.1

ERQA 45.8 27.3 42.5 39.8

HallusionBench 72.1 52.6 67.5 65.3

MMBench 87.9 54.2 85.2 84.9

MME 85.4 61.0 83.9 82.1

MMMU-Pro 26.4 11.9 24.1 18.5

MMRB2 54.2 48.3 52.6 49.5

MMStar 67.5 31.9 61.3 59.9

MMVP 82.7 54.0 81.3 77.0

NaturalBench 81.5 58.8 80.0 79.8

POPE 88.7 82.7 89.9 88.6

V* 76.4 35.6 76.4 73.8

VisualPuzzles 29.9 26.2 26.9 29.5

VisualWebBench 96.2 14.9 88.6 71.4

VL-RewardBench 69.3 48.1 68.8 65.0

Average70.242.867.765.4

Evaluation details

All four models are scored on the same items with the same scoring, so a few of d1-3B's and d1-omni-600M's numbers differ slightly from their own cards. d1-omni-600M has no Fast Decisions or board score. Average is the unweighted mean of the public benchmarks that all four models have, so it leaves out Fast Decisions and our board. Image benchmarks are read as decisions over each benchmark's options, on at most 1,000 rows each. Three of them are saturated: POPE, MME, and VisualWebBench.

Speed

Warm calls after model.compile() in bf16, median of 20 calls, and best of three runs on the H100 and the M5 Pro. Compiling takes about a minute on the first calls.

Hardware One question 3 questions 5K-token state 384 px picture Box Boxes 64 states in one call

NVIDIA H100 PCIe 5.7 ms 7.6 ms 52 ms 12.3 ms 16.1 ms 24.4 ms 1,132 / s

AMD MI325X 3.8 ms 5.6 ms 44 ms 8.7 ms 11.2 ms 16.5 ms 1,699 / s

Apple M5 Pro (mps) 18.6 ms 26.5 ms 753 ms 45.2 ms 53.5 ms 100 ms 109 / s

Citation

@article{liquidAI2026opend1,
  author  = {Liquid AI},
  title   = {Open d1: Edge decision models for text, vision, and audio},
  journal = {Liquid AI Blog},
  year    = {2026},
  note    = {https://www.liquid.ai/blog/open-d1},
}

@article{liquidAI2026LFM3,
  author  = {Liquid AI},
  title   = {Introducing LFM3: Every Modality for Specialized Intelligence},
  journal = {Liquid AI Blog},
  year    = {2026},
  note    = {https://www.liquid.ai/blog/lfm3},
}

For questions and feedback, join Discord. For custom deployment and fine-tuning support, contact Liquid AI.