Skip to content
Computer vision

Gemma 4 26B A4B

MoE-мультимодальность с 3,8B активных параметров

1 min readupdated

В репозитории Virtudant-AI выложена версия Gemma 4 26B A4B — мультимодальной модели Google DeepMind для текста и изображений. Это MoE-вариант: 25,2 млрд параметров всего и 3,8 млрд активных, контекст до 256K токенов. Карточка отсылает к гайду Unsloth по запуску и файнтюнингу Gemma 4.

Ключевые особенности

  • MoE-архитектура: 128 экспертов, 8 активных и 1 shared; при инференсе активна лишь часть параметров.
  • Понимает текст и изображения с поддержкой переменных соотношений сторон и разрешений; видео анализирует через последовательность кадров.
  • Контекст до 256K токенов; гибридное внимание чередует local sliding window и global attention, в глобальных слоях — unified Keys/Values и p-RoPE.
  • Configurable thinking mode и нативная поддержка system prompt.
  • Function calling, генерация/дополнение/исправление кода, 35+ языков «из коробки» после претрейна на 140+.
  • Лицензия Apache 2.0.

Пример

from transformers import AutoProcessor, AutoModelForCausalLM

MODEL_ID = 'google/gemma-4-26B-A4B-it'
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID, dtype='auto', device_map='auto'
)
ПодробнееMMLU Pro — 82.6%, AIME 2026 без инструментов — 88.3%, LiveCodeBench v6 — 77.1%.
Модель рассчитана на consumer GPU и рабочие станции; активных параметров 3.8B при 25.2B общих.
HLE без инструментов — 8.7% против 19.5% у 31B. Лицензия Apache 2.0.
Discuss Telegram 1 66

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 48 GB · ready files53 GB
  • FP8 / INT8recommendedweights 24 GB27 GB
  • 4-bitweights 14 GB15 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.

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.

    # Virtudant-AI/gemma-4-26B-A4B-it
    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 Virtudant-AI/gemma-4-26B-A4B-it --local-dir ~/models/gemma-4-26B-A4B-it
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to gemma-4-26B-A4B-it
  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. gemma-4-26B-A4Bfamily root
  2. gemma-4-26B-A4B-it
  3. gemma-4-26B-A4B-it

New posts about versions of gemma-4-26B-A4B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
Merges 5All on Hugging Face

About the model

Virtudant-AI/gemma-4-26B-A4B-it
Author
Virtudant-AI
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

661

Ask AI about this model

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

Comments

Language models

haimation

haimation