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.
Hy4-base
tencent·Base model·released 18 Nov 2026
vllm serve tencent/Hy4-baseHy4 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.
How much VRAM does Hy4-base need?
At Q4_K_M Hy4-base needs about 459 GB with an 8K context (459 GB at Q4, 1510 GB at full precision). Pick your GPU to see what fits and how fast it runs.
Needs: ≈459 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.
Advanced: context 8K · KV cache 16-bit
Smallest GPU for each quant and context (KV cache 16-bit)
| Quantisation | 8K | 16K | 32K | 64K | 128K | 1M |
|---|---|---|---|---|---|---|
| Q3_K_M | ||||||
| Q4_K_M | ||||||
| Q5_K_M | ||||||
| Q6_K | ||||||
| Q8_0 | ||||||
| FP16 |
≤16 GB card24–32 GB48–96 GBmore
| Quantisation | Total |
|---|---|
| Q2_KHeavy lossSmallest, noticeably worse quality | ≈250 GB |
| Q3_K_MNoticeable lossSmall, some quality loss | ≈370 GB |
| Q4_K_MrecommendedSmall lossRecommended default for local use | ≈459 GB |
| Q5_K_MNear-losslessNear-lossless for most tasks | ≈538 GB |
| Q6_KNear-losslessPractically lossless | ≈623 GB |
| Q8_0LosslessLossless in practice, twice the size of Q4 | ≈803 GB |
| FP16 / BF16LosslessFull precision, as released | ≈1510 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.
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?
At 8K context, Q4_K_M needs ≈459 GB(435 GB weights + 1.2 GB KV cache + 22 GB overhead). Pick a GPU above to see how much context fits.
Weights + overheadKV cache
Speed vs quality per quant
Smaller quants run faster and fit in less memory, but lose quality.
Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.
Real-world speed of Hy4-base
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-base locally
These commands load the original weights from Hugging Face.
vllm serve tencent/Hy4-baseservices:
vllm:
image: vllm/vllm-openai:latest
command: tencent/Hy4-base
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.
from transformers import pipeline
pipe = pipeline("image-text-to-text", model="tencent/Hy4-base", 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"])About Hy4-base
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-base
All alternatives →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.
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.
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…
Authors: Meta Superintelligence Labs Model Release Date: October 2026 License: Muse Model License
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.
Hy4-base FAQ
- How much VRAM does Hy4-base need?
- About 459 GB at Q4_K_M with an 8K context. Q8_0 needs about 803 GB and full precision about 1510 GB. Longer contexts need more memory for the KV cache.
- Can Hy4-base run on a consumer GPU?
- Not comfortably on a single one: it needs about 459 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-base 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-base?
- 1,048,576 tokens (1M). The full window needs far more memory than an 8K context.
- How do I run Hy4-base locally?
- With vLLM: vllm serve tencent/Hy4-base.
Rate this model
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.
0 comments