Skip to content
Computer vision

CoSTA-MIX

компактнее речь, живее диалог

1 min readupdated

CoSTA-MIX — чекпоинт для голосового диалога на китайском и английском, адаптированный из Step-Audio-2-mini. Модель на 8,89 млрд параметров сжимает последовательности речевых токенов обученным BPE-словарём, а перед синтезом волны разворачивает их обратно без потерь. Это финальная версия после расширения словаря, адаптации произношения и диалоговой адаптации.

Ключевые особенности

  • Базируется на Step-Audio-2-mini; поддерживает китайский и английский.
  • Речевые токены сжимаются через BPE; обратное развёртывание lossless до исходного speech-token space.
  • BF16-веса: 238 352 текстовых и речевых token ID, включая 80 000 добавленных speech tokens.
  • Отдельный speech BPE-словарь на 86 656 записей: 6 656 исходных и 80 000 merged units.
  • Apache-2.0; inference через кастомную реализацию Transformers 4.49.0 и репозиторий CoSTA, не через обычный pipeline.

Пример

export COSTA_LANGUAGE=mix
source scripts/env.sh
hf download D-YZ/CoSTA-MIX --local-dir "checkpoints/CoSTA-MIX"
hf download stepfun-ai/Step-Audio-2-mini \
  --include "token2wav/*" --local-dir "$BASE_MODEL_PATH"
export MODEL_PATH="$PWD/checkpoints/CoSTA-MIX"
CUDA_VISIBLE_DEVICES=0 bash inference/infer_costa.sh
ПодробнееБенчмарки в карточке не приведены — оценка на URO-Bench описана в статье.
Нужны Transformers 4.49.0, кастомная реализация и внешний token2wav-декодер из базовой модели.
Часть данных адаптации синтетическая; возможны фактические ошибки, качество зависит от языка и акустики.
Лицензия Apache-2.0.
Discuss Telegram 0 18

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 17 GB · ready files19 GB
  • FP8 / INT8recommendedweights 8.3 GB9.6 GB
  • 4-bitweights 4.7 GB5.6 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 3080 · BF16 — from RTX 4090

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.

    # D-YZ/CoSTA-MIX
    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 D-YZ/CoSTA-MIX --local-dir ~/models/CoSTA-MIX
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to CoSTA-MIX
  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 Step-Audio-2-mini (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 2All on Hugging Face
Fine-tunes 3All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

D-YZ/CoSTA-MIX
Author
D-YZ
Category
Computer vision
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

Computer vision

B6-27B

B6-27B