Skip to content
Other

d3-edge

компактная модель решений для роутинга запросов

2 min readupdated

d3-edge — это 0.59B мультимодальная фундаментальная модель принятия решений Decision 3.0 от vLLM Semantic Router, построенная на базе Qwen/Qwen3.5-0.8B. Она принимает текст или JSON, опционально изображения и видео, и отвечает на вопросы выбора, «да/нет» или оценки, возвращая вероятность для каждого варианта без генерации текста.

Ключевые особенности

  • Читает несколько изображений за запрос: PNG, JPEG или WebP — пути, URL, PIL или base64.
  • Понимает несколько видео за запрос: MP4, WebM, MOV или MKV, чтение на 2 кадра/с.
  • Поддерживает три типа решений: Choice, Yes/No и Score — все вопросы за один вызов.
  • Медианная задержка: 6.6 мс для текста, 58.6 мс с изображением и 274 мс для 10-секундного видео на AMD Instinct MI325X.
  • До 861 запроса/с с p99 ниже 1 с на одной MI325X через vLLM Semantic Router — 6.1× к AutoModel.
  • Лицензия Apache-2.0.

Пример

from transformers import AutoModel

model = AutoModel.from_pretrained("vllm-sr/d3-edge", trust_remote_code=True)
result = model.system_one(
    state="Клиент просит замену повреждённого товара",
    questions={"route": {"type": "choice", "instructions": "Кто должен обработать запрос?",
                         "criteria": {"returns": "Возвраты", "billing": "Оплата"}}},
)
print(result["answers"])
ПодробнееJev Decision Index 0.3.2 — 20.2 на официальном текстовом борде; public suite — 28.82.
+6.6 к Decision 2.0; Perception Test (video QA) — 46.6.
Пик памяти: AutoModel — 5.2 GiB; Engine mode — 15.9 GiB на реплику, 19.9 GiB с батчингом.
Видео читается с частотой 2 кадра/с, до 32 кадров.
Примеры
Discuss Telegram 0 22

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.1 GB · ready files1.7 GB
  • FP8 / INT8recommendedweights 0.6 GB1.1 GB
  • 4-bitweights 0.3 GB0.8 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # nam194/d3-edge
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download nam194/d3-edge --local-dir ~/models/d3-edge
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to d3-edge
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Qwen3.5-0.8B-Basefamily root
  2. Qwen3.5-0.8B
  3. d3-edge

New posts about versions of Qwen3.5-0.8B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 9All on Hugging Face
Merges 1All on Hugging Face

About the model

nam194/d3-edge
Author
nam194
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

220

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments