Skip to content
Other

Qwen Concept 27B

меньше размышлений, выше результаты

1 min read

Экспериментальная версия Qwen3.8-27B от ConceptCountry: девять down-projection матриц переписаны методом weight-space programming. Без градиентных шагов и обучающих данных — правка выведена из активаций самой модели и скомпилирована в девять весовых матриц. По карточке, модель меньше размышляет и набирает больше баллов.

Ключевые особенности

  • Девять down-projection переписаны через weight-space programming, без обучения и данных.
  • Thinking-токены снижены на 27–54% относительно немодифицированной базы.
  • GPQA Diamond +4.3 (интервал +1.8…+7.1), LiveCodeBench v6 +13.0 — по предзарегистрированному критерию.
  • Контроль: случайная правка с той же нормой, слоями и амплитудой, но другим направлением не даёт эффекта; интервалы на всех бенчмарках пересекают ноль.
  • Форматы: bf16 ~52 ГБ, FP8_BLOCK ~30 ГБ, NVFP4 W4A16 ~20 ГБ, GGUF Q8_0 ~27 ГБ (файл бенчмарков) и Q4_K_M ~16 ГБ (квантование от bf16, не от Q8_0).
  • В каждом формате есть SHA256SUMS; статус — экспериментальная, на оценке.

Пример

HF_ENDPOINT=https://models.concept.country hf download ConceptCountry/qwen-concept-27b
ПодробнееБенчмарки: GPQA Diamond +4.3 [+1.8, +7.1], LiveCodeBench v6 +13.0, thinking-токены −27–54%.
Все цифры получены на bf16-весах, обслуживаемых как Q8_0; кванты — удобство деплоя, не отдельно оценённые артефакты.
NVFP4 требует железа класса Blackwell; эффект 4-битных квантов отдельно не проверялся.
Статус: экспериментальная, на оценке. Лицензия Apache-2.0.
Discuss Telegram 1

Will it run?

Runs in GGUF Q4_K_M

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

  • GGUF Q8_0recommendedweights 27 GB · ready files30 GB
  • GGUF Q4_K_Mweights 16 GB · ready files18 GB

Tip: go 8-bit.GGUF Q8_0 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: GGUF Q8_0 — from RTX 5090

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.

    # ConceptCountry/qwen-concept-27b
    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 ConceptCountry/qwen-concept-27b --local-dir ~/models/qwen-concept-27b
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to qwen-concept-27b
  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

New posts about versions of Qwen3.8-27B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

ConceptCountry/qwen-concept-27b
Author
ConceptCountry
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

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

Comments