selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

tencent logo

Hy4

tencent·Instruct·released 29 Oct 2026

tencent-hy-license774BA53.6BMoE1M ctx ≈461 GB
Your GPU
vllm serve tencent/Hy4
All 2 runners ↓

Hy4 is a new-generation Mixture-of-Experts (MoE) flagship model developed by the Tencent Hy Team, supporting text, image, and video understanding. The language backbone comprises 770B total parameters, of which 49B are activated per token.

Notify me when

How much VRAM does Hy4 need?

At Q4_K_M Hy4 needs about 461 GB with an 8K context (461 GB at Q4, 1516 GB at full precision). Pick your GPU to see what fits and how fast it runs.

--vram
any
--gpus
--quant

Needs: ≈462 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.

Change quant, context or KV cache to compare
Advanced: context 8K · KV cache 16-bit
Context
KV cache

Smallest GPU for each quant and context (KV cache 16-bit)

Quantisation8K16K32K64K128K1M
Q3_K_M
Q4_K_M
Q5_K_M
Q6_K
Q8_0
FP16

≤16 GB card24–32 GB48–96 GBmore

QuantisationQualityWeightsKV cacheTotal
Q2_KHeavy lossSmallest, noticeably worse qualityHeavy loss237 GB est.1.2 GB≈251 GB
Q3_K_MNoticeable lossSmall, some quality lossNoticeable loss353 GB est.1.2 GB≈372 GB
Q4_K_MrecommendedSmall lossRecommended default for local useSmall loss438 GB file1.2 GB≈462 GB
Q5_K_MNear-losslessNear-lossless for most tasksNear-lossless513 GB est.1.2 GB≈540 GB
Q6_KNear-losslessPractically losslessNear-lossless594 GB est.1.2 GB≈626 GB
Q8_0LosslessLossless in practice, twice the size of Q4Lossless766 GB est.1.2 GB≈806 GB
FP16 / BF16LosslessFull precision, as releasedLossless1443 GB est.1.2 GB≈1516 GB

Estimates for 8K tokens of context: weights + KV cache + runtime overhead. All experts stay in memory, but only the active ones are read per token. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.

How to pick a quant

Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.

  • LosslessF16, Q8_0
  • Near-losslessQ6_K, Q5_K_M
  • Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
  • Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
  • Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K

Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B

How much context fits?

Model quant
KV cache

At 8K context, Q4_K_M needs ≈462 GB(438 GB weights + 1.2 GB KV cache + 22 GB overhead). Pick a GPU above to see how much context fits.

02004006008001K2K4K8K16K32K64K128K256K512K1MGB

Weights + overheadKV cache

Speed vs quality per quant

Smaller quants run faster and fit in less memory, but lose quality.

Q2_KQ3_K_MQ4_K_MQ5_K_MQ6_KQ8_0F16

Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.

Real-world speed of Hy4

Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →

No medians yet. Be one of the first 3 to report a setup.

All reports

Run Hy4 locally

These commands load the original weights from Hugging Face.

Serve an OpenAI-compatible APIdocs
vllm serve tencent/Hy4
Docker Compose (NVIDIA)docs
yaml
services:
  vllm:
    image: vllm/vllm-openai:latest
    command: tencent/Hy4
    ports:
      - "8000:8000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

API on port 8000.

Install vLLM

Ask about an imagedocs
python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="tencent/Hy4", dtype="auto", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
            {"type": "text", "text": "Describe this image."},
        ],
    }
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])

Install Transformers

  • MLXcommunity · library

Quantised versions and downloads

RepositoryFormatQuantSizeDownloads
AngelSlim/Hy4-GGUFGGUFIQ2332 GB28
AngelSlim/Hy4-GGUFGGUFQ4_K_M470 GB28
Xylolabs/Hy4-GGUFGGUF———
Xylolabs/Hy4-MLX-4bitMLX4BIT——
Xylolabs/Hy4-MLX-8bitMLX8BIT——
nace-ai/Hy4-MLX-4.6bitMLX———
Jundot/Hy4-oQ4h-mtpMLX———

About Hy4

Hy4 is a new-generation Mixture-of-Experts (MoE) flagship model developed by the Tencent Hy Team, supporting text, image, and video understanding. The language backbone comprises 770B total parameters, of which 49B are activated per token. It consists of 78 layers, with a standard dense FFN in the first layer and MoE FFNs in the remaining 77 layers. Each MoE layer contains 256 routed experts and 1 shared expert; every token activates the top-8 routed experts along with the shared expert. In addition to the backbone, a 5-layer DFly2 drafter (3.3B parameters) is built in for speculative decoding. On the architecture side, inspired by DeepSeek and GLM, the attention module employs Gated DeepSeek Sparse Attention (Gated DSA) with IndexCache for cross-layer sparse index reuse. The residual pathway uses iHC (identity Hyper-Connections) to expand inter-layer information flow. A ViT supports dynamic image resolutions, with an adapter converting image and video frame features into visual tokens for the language model to process alongside text tokens. DFly2, a parallel drafter, reuses Hy4's KV cache and sparse indices to support long-context speculative decoding at a low, constant cost. Bits-per-byte (BPB) on internal code & agent validation, lower is better: Few-shot pretrain benchmark evaluation:

Full model card on Hugging Face →

Alternatives to Hy4

All alternatives →
deepseek-ai logo

DeepSeek-V4.1-Flash

deepseek-ai·Vision

≈454 GB

We introduce DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts (MoE) model with 552B backbone parameters and support for contexts of up to one million tokens. The model natively processes images and text, and generates text autoregressively.

763BA188B1M ctx MIT
1.4MSept 2026gguf
deepseek-ai logo

DeepSeek-V4.1-Flash-1017

deepseek-ai·Vision

≈454 GB

DeepSeek-V4.1-Flash-1017 is a minor update to DeepSeek-V4.1-Flash, delivering substantial improvements across reasoning, coding, agentic tasks, and visual benchmarks.

763BA188B1M ctx MIT
505.4KOct 2026gguf
thinkingmachines logo

Inkling

thinkingmachines·Vision

≈567 GB

Inkling is a general-purpose multimodal model that accepts text, image and audio inputs and generates text outputs. It is intended for use in English and other languages, and across multiple coding languages.

952BA18.1B1M ctx Apache 2.0
1.3MJul 2026gguf
moonshotai logo

Kimi-K2.7-Code

moonshotai·Vision

≈632 GB

Kimi K2.7 Code is a coding-focused agentic model built upon Kimi K2.6. With substantial improvements on real-world long-horizon coding tasks, it strengthens end-to-end task completion across complex software engineering workflows while improving token efficiency, reducing thinking-token usage by…

1.03TA33.3B256K ctx modified-mit
324.1KJun 2026ollama
meta-models logo

Muse-Spark-Open-816B-A42B

meta-models·Vision

≈490 GB

Authors: Meta Superintelligence Labs Model Release Date: October 2026 License: Muse Model License

822BA42B1M ctx muse-model-license
14.4KSept 2026gguf
moonshotai logo

Kimi-K2.6

moonshotai·Vision

≈632 GB

Kimi K2.6 is an open-source, native multimodal agentic model that advances practical capabilities in long-horizon coding, coding-driven design, proactive autonomous execution, and swarm-based task orchestration.

1.03TA33.3B256K ctx modified-mit
198.9KApr 2026ollama

Hy4 FAQ

How much VRAM does Hy4 need?
About 461 GB at Q4_K_M with an 8K context. Q8_0 needs about 806 GB and full precision about 1516 GB. Longer contexts need more memory for the KV cache.
Can Hy4 run on a consumer GPU?
Not comfortably on a single one: it needs about 461 GB at Q4. Split it over several GPUs, use a large unified-memory machine, or offload part of it to system RAM.
Can I use Hy4 commercially?
Unclear. It is released under tencent-hy-license. Custom licence: read the model card before using it commercially. Check the licence on the model card before using it in a product.
What is the context length of Hy4?
1,048,576 tokens (1M). The full window needs far more memory than an 8K context.
How do I run Hy4 locally?
With vLLM: vllm serve tencent/Hy4.

Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.