Skip to content

Hy4 Flash VRAM Requirements

Tencent, China · Released November 17, 2026New

Hy4 Flash has 220.3B parameters (11.3B active per token). In BF16 its weights alone take 409 GiB; quantized to 4-bit, about 123 GiB. Each 32K-token request adds 0.45 GiB of KV cache. Cheapest way to run it today: 4 × GeForce RTX 3090 (24 GB) at $0.436 per hour. It reads text, images and video, calls tools and lets you set its reasoning effort.

Size your own setup in the calculator →

Hy4 Flash at a glance

Made by
Tencent, China
Released
November 17, 2026
Parameters
220.3B, 11.3B active
Context window
1M tokens
License
Tencent Hy License license
Reads
Text, images and video
Tool calling
Supported
Reasoning
Adjustable reasoning effort
Runs with
SGLang, MLX (Mac) · vLLM recipe (opens in a new tab)

GGUF builds exist, such as peasantsmith/Hy4-Flash-Maya-GGUF (opens in a new tab), but they need a patched llama.cpp: hy_v4_flash is not in mainline yet

VRAM by precision and context length

Weights plus the KV cache of one request (BF16 cache), in GiB. Add 1 to 3 GiB for the inference engine itself.

Weights formatWeights only+ 8K context+ 32K context+ 128K context+ 1M context
FP8 (native)266267267268280
BF16 / FP16409409409410423
FP8 / INT8205205205207219
4-bit (GGUF Q4_K_M)123123124125137

Hy4 Flash GGUF files

Exact size of each quantization in AngelSlim/Hy4-flash-GGUF (opens in a new tab), and the smallest GPU that runs it with llama.cpp, Ollama or LM Studio and an 8K-token context.

QuantizationFile sizeSmallest GPU
Q4_K_M157 GiBMore than 96 GB: see the calculator
Q3_K145 GiBMore than 96 GB: see the calculator
IQ2_AH126 GiBRTX PRO 6000 Blackwell (96 GB)

Read every day from Hugging Face. Split files are added together; vision projectors are left out.

How many GPUs to run Hy4 Flash

Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.

GPUFP8 (native)BF16 / FP168-bit4-bit
GeForce RTX 4090 24 GBNot on this GPU16 (several machines)74
GeForce RTX 5090 32 GBNot on this GPU16 (several machines)53
RTX PRO 6000 Blackwell 96 GBNot on this GPU421
Mac M5 Max 128 GB 96 GB usableNot on this GPU421
Ryzen AI Max+ 395 128 GB 96 GB usableNot on this GPU421
DGX Spark (GB10) 128 GBNot on this GPU321
NVIDIA H100 80 GB3532
NVIDIA H200 141 GB2321
NVIDIA B200 (HGX) 180 GB2211

8-bit: FP8 with vLLM (INT8 on GPUs without FP8), Q8_0 with llama.cpp. 4-bit: AWQ / GPTQ with vLLM, Q4_K_M with llama.cpp, both counted at the Q4_K_M size. Not on this GPU: the official FP8 weights do not run there (llama.cpp needs a GGUF version, and vLLM needs a GPU with FP8 support).

Cheapest way to run Hy4 Flash today

For each GPU: the fewest cards that fit Hy4 Flash with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-11-22).

On-demand prices from AWS, Google Cloud, Azure, RunPod, Vast.ai, Oracle Cloud, OVHcloud, Scaleway, Verda, Akash and UpCloud, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.

AWS, Google Cloud and UpCloud prices: Spare Cores (opens in a new tab), CC BY-SA 4.0 (opens in a new tab), shared under the same license.

Recommended settings for Hy4 Flash

Sampling defaults from the official generation_config.json (opens in a new tab) of tencent/Hy4-flash.

SettingValue
Temperature0.9
Top-p1

vLLM and SGLang apply these defaults unless a request sets its own. With llama.cpp, pass them yourself: --temp 0.9 --top-p 1

Reasoning. Adjustable reasoning effort. Reasoning tokens take context and KV cache like any other token: leave room for them.

KV cache: how memory grows with context

Architecture: MoE 512 latent experts + 1 shared · MLA (512 + 64) on 42 layers: 12 sparse-attention layers (indexer keys per 4-token block) + 30 sliding-window layers (512 tokens) · 63B Engram tables in RAM · image input. Its KV cache is 3.3× smaller than if every layer used full attention, because it keeps only a short window of tokens at full detail and compresses the rest of the sequence. Only 11.3B parameters are active per token, but all 220B must sit in memory: VRAM depends on the total, speed on the active part.

Context per requestBF16 cacheFP8 cache
8K tokens0.12 GiB0.07 GiB
32K tokens0.45 GiB0.26 GiB
128K tokens1.75 GiB1.02 GiB
1M tokens13.9 GiB8.08 GiB

Hy4 Flash VRAM FAQ

How much VRAM does Hy4 Flash need?

In BF16 the weights alone take 409 GiB (439 GB). In FP8 that is 205 GiB, and about 123 GiB with 4-bit quantization (Q4_K_M); the official FP8 checkpoint is 266 GiB. Each request then adds KV cache: 0.12 GiB at 8K tokens and 0.45 GiB at 32K (BF16 cache).

Can Hy4 Flash run on a single RTX 4090 (24 GB)?

No. Even in 4-bit it needs 4 RTX 4090s. In 4-bit it fits on a single RTX PRO 6000 Blackwell (96 GB).

Can Hy4 Flash run on an RTX 5090 (32 GB)?

No. Even in 4-bit it needs 3 RTX 5090s.

Can Hy4 Flash run on a Mac?

Yes. In 4-bit (Q4_K_M with llama.cpp) it needs a Mac with 128 GB of unified memory or more, of which macOS gives the GPU about 96 GB, for one request with an 8K-token context. In 8-bit (Q8_0) it needs 256 GB or more.

Can Hy4 Flash run on a DGX Spark or a Ryzen AI Max+ 395 (128 GB)?

Yes. In Q4_K_M (123 GiB with an 8K-token context) it fits on a DGX Spark (128 GB) and on a Ryzen AI Max+ 395 with 128 GB, whose GPU gets up to 96 GB, with llama.cpp and one request.

What is the cheapest way to run Hy4 Flash?

On 2026-11-22, the cheapest on-demand setup is 4 × GeForce RTX 3090 (24 GB) with Q4_K_M weights on Vast.ai, at $0.436 per hour (about $318.28 per month). Next: 7 × RTX A4000 (16 GB) with Q4_K_M weights on Vast.ai, at $0.616 per hour. Sized for one 8K-token request; prices are checked every hour.

How many H100 GPUs does Hy4 Flash need?

With one request and an 8K-token context: 5 in BF16, 3 in FP8 and 2 in 4-bit (vLLM, 90% of memory usable). Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.

Can Hy4 Flash read images?

Yes. Hy4 Flash reads text, images and video. Images add tokens to the context, so they also add KV cache.

Does Hy4 Flash support tool calling?

Yes, Hy4 Flash supports tool (function) calling: it can ask your application to run a function and use the result.

Does Hy4 Flash have a thinking mode?

Adjustable reasoning effort. Reasoning tokens count toward the context and the KV cache.

What temperature should I use with Hy4 Flash?

The official defaults are temperature 0.9, top-p 1.

Hy4 Flash VRAM badge

For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/hy4-flash.json.

Hy4 Flash VRAM badge

[![Hy4 Flash VRAM](https://studiotvai.com/badge/hy4-flash.svg)](https://studiotvai.com/vram-requirements/hy4-flash)

Other models

Same family (large moe): MiniMax M3.1 Flash · MiniMax M3 · MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · Atria Dawn · Hydrogen-1.0 · Atria Dawn Preview · Hy4 Preview · K-EXAONE 2.0 750B-A37B · Inkling · Inkling 2 Small · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · LongCat-2.0 · LongCat-2.5 · ERNIE 5.5 Flash Base · MiMo-V2.6 Pro · MiMo-V2.6 Flash · Muse Spark Open · Beam 1.0 501B-A23B · Kimi K3 · Mistral Large 4 · Mistral Large 3 · Mistral Small 4 119B · Leanstral 1.5 119B-A6B · Mixtral 8x22B · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Intern S2 397B · DeepSeek V3 · DeepSeek V3.1 · Kimi K2.7 Code · Kimi K2.5 · Kimi K2 · GLM-5.2 · GLM-5.1 · GLM-5 · MiniMax M2.5 · MiniMax M2 · MiMo-V2.5 · Ornith-1.0 397B · Qwen3 235B-A22B · Qwen3.5 397B-A17B · GLM-4.7 · GLM-4.5 · Ling-3.1-flash · Step 5 Preview · Hy3 · Kimi K3.1 Pro · DeepSeek V3.2 · Laguna M.1 · Qwen3-VL 235B-A22B · Nex-N2.6 · Sage-S2.1 · Hy3-preview · Hy4

All 233 other models

Llama 3.1 8B · Llama 3 8B · Llama 3.2 3B · Nanbeige4.3 33B · Granite 4.2 30B · Granite 4.2 8B · Granite 4.2 3B · MiniCPM5 2B · MiniCPM5 1B · Ministral 3 14B · Ministral 3 8B · Ministral 3 3B · Devstral 2 123B · Devstral Small 2 24B · Mistral Medium 3.5 128B · Magistral Small 1.2 24B · Mistral Small 3 24B · Mistral Large 2411 123B · Pixtral Large 124B · Codestral 22B · Mistral Small 2409 22B · Pixtral 12B · Nemotron 3.5 Parse · Gemma 2 9B · Gemma 3 4B · Gemma 3 12B · Gemma 3 27B · LLaVA 1.6 Mistral 7B · Mistral Small 3.2 24B · Qwen3 32B · Llama 3.3 70B · Llama 3.1 70B · Llama 3.1 405B · Hy-Memory 30B-A3B · Kimi VL A3B Instruct · Mixtral 8x7B · Unlimited-OCR · Unlimited-OCR 2 · Xing4.0 29B-A4B · Qwen3.5 4B · MiniCPM-V-4.7 1B · Raon-1 mini preview · Raon-1 nano · Qwen3.5 9B · Qwen3.8 27B · Qwen3.6 35B-A3B · Maxwell V1 Flash · Clover 35B-A3B Preview · SenseNova Loop 9B Preview · Qwen3.5 122B-A10B · Qwen3.8-Flash-Next · Qwen3.8-Flash-Next E384 FP4 V2 · LFM3 1.7B · LFM3-VL 2B · d1 2B · LFM2.5 8B-A1B · LFM2 24B-A2B · Gemma 4 12B · Gemma 4 31B · Gemma 4 26B-A4B · gpt-oss 20B · gpt-oss 120B · Llama 4 Scout 17B-16E · Llama 4 Maverick 17B-128E · Kolibri-1 · Recursive Flint · Nemotron Labs Dory 1.0 4B · Step 3.5 Flash · Step 3.7 Flash · Laguna S 2.1 · Laguna XS 2.1 · Command A+ · Inkling Small · Muse Glimmer 30B · Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · Nemotron 3 Ultra 550B-A55B · Mamba-Codestral 7B · Nemotron 3.5 Super VL 120B-A12B · Nemotron 3.5 Bolt 74B-A9B · MiniMax M3.1 Flash · MiniMax M3 · MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · Atria Dawn · Hydrogen-1.0 · Atria Dawn Preview · Hy4 Preview · K-EXAONE 2.0 750B-A37B · Inkling · Inkling 2 Small · Inkling 2 Mini · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · LongCat-2.0 · LongCat-2.5 · ERNIE 5.5 Flash Base · MiMo-V2.6 Pro · MiMo-V2.6 Flash · Muse Spark Open · Beam 1.0 501B-A23B · Kimi K3 · Mistral Large 4 · Mistral Large 3 · Mistral Small 4 119B · Leanstral 1.5 119B-A6B · Mixtral 8x22B · Qwen3-Coder-Next 80B-A3B · Qwen3-Coder 30B-A3B · Qwen3-Coder 480B-A35B · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Ornith 1.5 35B-A3B · Ornith 1.5 9B · Intern S2 397B · Clef · Clef Flash · d1 3B · Humanizer · JEV 27B VL · Mellum2.1 12B-A2.5B · LightOnOCR-3 4B · Spark-X2.5 4B · Agnes-3.0-Qwen · Youtu-Parsing-Omni · Agnes-3.0-Flash · Mistral 7B v0.3 · Mistral Small 3.1 24B · Mistral Nemo · Ministral 8B · DeepSeek V3 · DeepSeek R1 Qwen3 8B · DeepSeek R1 Distill Qwen 32B · DeepSeek V3.1 · DeepSeek R1 Distill Qwen 14B · DeepSeek R1 Distill Qwen 7B · DeepSeek R1 Distill Llama 8B · Qwen3 4B · Qwen3 8B · Qwen2.5 7B · Qwen3 14B · Qwen2.5 3B · Qwen3.6 27B · Qwen3.5 27B · Qwen2.5-Coder 7B · Qwen2.5 32B · Qwen2.5-Coder 14B · Qwen3.5 35B-A3B · Qwen3 30B-A3B · Qwen2.5 14B · Qwen2.5-Coder 32B · Kimi K2.7 Code · Kimi K2.5 · Kimi K2 · GLM-5.2 · GLM-5.1 · GLM-5 · MiniMax M2.5 · MiniMax M2 · MiMo-V2.5 · Ornith-1.0 397B · Qwen2.5-Coder 3B · Qwen3 235B-A22B · Qwen3.5 397B-A17B · Gemma 4 E4B · Gemma 4 E2B · Nemotron 3 Nano 4B · Nemotron 3.5 Lightning 30B-A3B · MiniCPM-V-4.7 35B-A3B · LFM2.5 24B-A2B · GLM-4.7 · Qwen3-VL 8B · Qwen2.5-VL 7B · Qwen3-VL 4B · Qwen2.5-VL 3B · Qwen2.5-VL 32B · Qwen2-VL 7B · GLM-4.5-Air · GLM-4.5 · phi-4 · Phi-3.5-mini · Phi-4-mini · Phi-3-mini-128k · Ling-3.1-flash · Step 5 Preview · Hy3 · Ornith-1.5 122B · Spark-X2.6 4B · Sage-A1.5 · MLExpert · Agents-A1 4B · Ling-3.0-flash-Sante · MiniCPM-OCR-1.0 · WeVisOne-VL 4B · Granite 4.1 3B · Qwen3-VL 30B-A3B · Mage-VL-Cowork · Spark-X2.6 1.7B · Qwen3.5 4B-Qwenputer · Kimi K3.1 Pro · 6.8-Code2Video · Qwen3.5 4B-GLB-3D-Coder · DeepSeek V3.2 · GLM-4.7-Flash · Laguna M.1 · Laguna XS.2 · Qwen3-VL 235B-A22B · Qwen2 7B · Qwen2.5 72B · Qwen3-VL 32B · Qwen2.5 14B-1M · Qwen3-Next 80B-A3B · Qwen2.5-VL 72B · Nemotron Nano 9B v2 · Qwen3.5 0.8B-weight-space-programming · Axion-1.0 35B · Hermes-3-Llama-3.1 8B · Qwen3.8 27B-DSpine-b16 · Granite 3.3 8B · MNN-Pocket 0.3B · S5-nano · Nex-N2.6 · Sage-S2.1 · KORMo-Web 11B · Hy3-preview · ERNIE 4.5 21B-A3B · Aion-1.0 2.3B · Hy4

Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.