Skip to content
Other

Hy4-Flash: A 220B Multimodal Model by Tencent Hunyuan

2 min read

Hy4-Flash is a large-scale general-purpose multimodal model developed by Tencent Hunyuan and adapted for the FlagOS ecosystem to run on diverse hardware platforms. The model features approximately 220 billion total parameters (including around 64 billion Engram parameters) and activates roughly 12 billion parameters per token. It supports text and image processing, a context length of up to 1 million tokens, and a hybrid architecture featuring Latent MoE (Mixture of Experts). The model is optimized for office scenarios, including deep search, document processing, and presentation generation, delivering reduced KV Cache memory consumption and high throughput.

Key features

  • Native support for text and visual inputs with a context length of up to 1 million tokens.
  • Latent MoE and Engram architecture, reducing KV Cache size by six times compared to the base version of Hy4.
  • Adaptation for the FlagOS stack for cross-hardware deployment on various chips (including Moore Threads, NVIDIA, and others).
  • Strong benchmark results (for example, 96 on Math-500 and 92.6 on mmbench).

What you can do

  • Deploy a local inference server for long-context processing and multimodal tasks.
  • Use the model for workflow automation, deep search, and presentation (PPT) generation.
  • Integrate the model into your workflows using ready-to-use deployment scripts and FlagOS Docker containers.

Example

An example request to a running inference server via the API using curl:

curl --fail-with-body --silent --show-error --max-time 180 \
  [http://127.0.0.1:30200/v1/chat/completions](http://127.0.0.1:30200/v1/chat/completions) \
  -H 'Content-Type: application/json' \
  --data-binary '{"model":"Hy4-flash","messages":[{"role":"user","content":"只回答数字:1+1等于多少?"}],"temperature":0,"max_tokens":32,"stream":false,"chat_template_kwargs":{"reasoning_effort":"no_think"}}'
Read moreDetails
Parameters: 220.25 billion total parameters, ~12B active.
Benchmarks: Math-500 — 96, GPQA_Diamond — 92.42, mmbench — 92.6, mmstar — 80.93.
License: Apache License 2.0.
Requirements: Supports distributed inference with tensor parallel size (tp-size) 8.
Discuss Telegram 0 2

Will it run?

Won't run

Not enough memory even for 4-bit. Below are GPUs you can rent.

  • I8recommendedweights 268 GB · ready files296 GB
  • 4-bitweights 115 GB127 GB

Tip: go 8-bit.I8 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

No single instance holds the whole model: it needs RAM offload or a cluster.

Your own GPU: Won't fit any home GPU — renting is cheaper.

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.

    # FlagRelease/Hy4-flash-INT8-mthreads-FlagOS
    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 FlagRelease/Hy4-flash-INT8-mthreads-FlagOS --local-dir ~/models/Hy4-flash-INT8-mthreads-FlagOS
    # 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 Hy4-flash-INT8-mthreads-FlagOS (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

FlagRelease/Hy4-flash-INT8-mthreads-FlagOS
Author
FlagRelease
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

20

Ask AI about this model

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

Comments