Skip to content
Other

Nemotron-3.5-Embed-1B-BF16

поиск по тексту и страницам документов

2 min read

NVIDIA представила Nemotron-3.5-Embed-1B-BF16 — мультимодальный эмбеддер на 1,55 млрд параметров для retrieval-задач. Это bi-encoder с Pixtral vision encoder и языковой частью от Nemotron-3-Embed-1B: документ можно подать как изображение, текст или их комбинацию, а запрос — только текстом. На выходе — 2048-мерный вектор для семантического поиска и RAG.

Ключевые особенности

  • Мультимодальный вход: страницы с текстом, таблицами, графиками и инфографикой — как image, text или image + text.
  • Мультиязычность: текстовый retrieval оценён на 34 языках, image retrieval — на 18 языках.
  • Контекст до 32 768 токенов; изображение по умолчанию до 1 120 px по длинной стороне, можно увеличить до 1 540 px.
  • Эмбеддинг 2048 измерений; поддерживается усечение до 1024 или 512 измерений с L2-нормализацией.
  • Дистилляция из Nemotron-3.5-Embed-8B; текстовое пространство совпадает с Nemotron-3-Embed-1B-BF16, добавлен image retrieval.
  • Разрешено коммерческое и некоммерческое использование по лицензии OpenMDW-1.1.

Пример

from sentence_transformers import SentenceTransformer
model = SentenceTransformer("nvidia/Nemotron-3.5-Embed-1B-BF16", device="cuda", model_kwargs={"dtype": "bfloat16", "attn_implementation": "flash_attention_2"})
query = "What is NVIDIA NeMo Framework?"
docs = ["NeMo Framework offers comprehensive functionalities...", {"image": "https://docs.nvidia.com/nemo-framework/user-guide/24.07/_images/nemo-llm-mm-stack.png"}]
qe = model.encode_query([query])
de = model.encode_document(docs)
print(model.similarity(qe, de))
ПодробнееСреди моделей сопоставимого размера заявлен SOTA на мультиязычных и мультимодальных retrieval-бенчмарках; конкретных цифр в карточке нет.
1,55B параметров, BF16; примеры рассчитаны на CUDA GPU и FlashAttention 2. Отдельных требований к VRAM не указано.
Максимальный контекст — 32 768 токенов, включая текст и image tokens.
Лицензия OpenMDW-1.1; базовые Ministral-3-3B/8B — Apache 2.0.
Discuss Telegram 0 18

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 2.9 GB · ready files3.7 GB
  • FP8 / INT8recommendedweights 1.4 GB2.1 GB
  • 4-bitweights 0.8 GB1.4 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.

    # on1onmangoes/Nemotron-3.5-Embed-1B-BF16
    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 on1onmangoes/Nemotron-3.5-Embed-1B-BF16 --local-dir ~/models/Nemotron-3.5-Embed-1B-BF16
    # The model is in ~/models. How to run it is in the model card (link above).
  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. Ministral-3-3B-Base-2512family root
  2. Ministral-3-3B-Instruct-2512-BF16
  3. Nemotron-3.5-Embed-1B-BF16

New posts about versions of Ministral-3-3B-Base-2512 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 7All on Hugging Face

About the model

on1onmangoes/Nemotron-3.5-Embed-1B-BF16
Author
on1onmangoes
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

180

Ask AI about this model

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

Comments