Skip to content
Audio

Apriel Voice 27B Heretic

голосовая модель без цензуры

1 min readupdated

Это decensored-версия Apriel Voice 27B Preview от ServiceNow, сделанная через Heretic v2.0.0dev методом ARA. Модель на 26,49 млрд параметров понимает речь, рассуждает по контексту и генерирует текст или tool calls на общей backbone Gemma 4; синтез речи остаётся внешнему TTS.

Ключевые особенности

  • Отказы снижены до 4/100 против 99/100 у оригинала; KL divergence — 0.1818.
  • Архитектура: Whisper large-v3-turbo speech encoder + learned projector + Gemma 4 26B-A4B backbone.
  • Умеет транскрипцию, перевод речи и ответы на голосовые вопросы; принимает аудио, текст и изображения.
  • Генерирует structured tool calls и следует многошаговым инструкциям; инструменты выполняет приложение.
  • Подходит для голосовых ассистентов и support-сценариев: текст ответа + внешний синтез и turn-taking на стороне приложения.
  • Не full-duplex: чекпойнт выдаёт текст и tool calls, а не нативную речь.

Пример

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")
r = client.chat.completions.create(
    model="Apriel-Voice-27B",
    messages=[{"role":"user","content":[
        {"type":"audio_url","audio_url":{"url":"..."}},
        {"type":"text","text":"Transcribe this speech."}]}])
print(r.choices[0].message.content)
ПодробнееОтказы: 4/100 против 99/100 у оригинала; KL divergence — 0.1818.
Сервинг: vLLM v0.32.0, примеры на 4 GPU; оптимизированный образ даёт ~2× throughput на 4×H100 80 GB.
Ограничения: decensoring неполный — советуют менять system prompt; аудио 16 kHz mono, качество на разных языках неравномерное.
Длинные записи: хорошие результаты до ~30 минут, дальше качество может падать.
Примеры
Discuss Telegram 0 26

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

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.

    # lodrick-the-lafted/Apriel-Voice-27B-heretic
    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 lodrick-the-lafted/Apriel-Voice-27B-heretic --local-dir ~/models/Apriel-Voice-27B-heretic
    # 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

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

No known relatives of this model yet. All derivatives on Hugging Face

About the model

lodrick-the-lafted/Apriel-Voice-27B-heretic
Author
lodrick-the-lafted
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

260

Ask AI about this model

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

Comments