LLMBOTTLENECK.COM
Tencent / hy_v3

Hy3-preview

298.8 billion parameters, routing 8 of 192 experts per token.

Open in the calculator →

Architecturepublished data
Quick answer
Hy3-preview needs about 181.2 GB at Q4_K_M — the format most people download — with 8,192 tokens of context. It fits entirely on 1 of 18 common devices, starting with the Apple M2 Ultra (192 GB). What else fits in 192 GB. No hardware for it? Rent a GPU that runs it, priced live.

Sized at 8,192 tokens of context with the whole model in device memory, across 18 common devices. Speeds are estimates, not benchmarks.

Find the Apple M2 Ultra, the smallest card listed here that holds it: Amazon ↗ · eBay (new and used) ↗Store links may pay us a commission. They never decide which card is suggested — the memory arithmetic does.

Weight-file size by format

FormatSizeRangeBasis
FP16587.2 GB574.0 GB – 600.5 GBreconstructed size
Q8_0312.1 GB305.0 GB – 319.1 GBreconstructed size
Q6_K241.0 GB235.6 GB – 246.4 GBreconstructed size
Q5_K_M208.4 GB203.7 GB – 213.1 GBreconstructed size
Q5_0202.1 GB197.5 GB – 206.6 GBreconstructed size
Q4_K_M177.7 GB173.7 GB – 181.7 GBreconstructed size
Q4_0165.5 GB161.7 GB – 169.2 GBreconstructed size
Q3_K_M145.2 GB132.5 GB – 157.9 GBreconstructed size
Q2_K106.5 GB100.4 GB – 112.7 GBreconstructed size

A file size is not the memory a run needs: the KV cache and the runtime reserve come on top, and the calculator adds both.

What the labels mean
published data
Read from a published source — a file's byte count, a model's configuration or a manufacturer's specification — or exact arithmetic on such values. Not a measurement on a machine.
reconstructed size
Weight size reconstructed from the pinned architecture, because no published file exists.
size range
Only a lower and an upper bound are claimed for this weight size.

Hardware ladder

Q4_K_M at 8,192 tokens · 1 of 18 devices hold it entirely
DeviceMemoryVerdictDecodeCalculator
NVIDIA GeForce RTX 5070181.2 GB of 12 GB77 layers on system RAM~5.3 tok/s with offloadOpen →
AMD Radeon™ RX 9070 XT181.2 GB of 16 GB75 layers on system RAM~5.4 tok/s with offloadOpen →
NVIDIA GeForce RTX 4080181.2 GB of 16 GB75 layers on system RAM~5.4 tok/s with offloadOpen →
NVIDIA GeForce RTX 5080181.2 GB of 16 GB75 layers on system RAM~5.4 tok/s with offloadOpen →
NVIDIA GeForce RTX 5070 Ti181.2 GB of 16 GB75 layers on system RAM~5.4 tok/s with offloadOpen →
NVIDIA GeForce RTX 5060 Ti 16GB181.2 GB of 16 GB75 layers on system RAM~5.3 tok/s with offloadOpen →
NVIDIA GeForce RTX 4090181.2 GB of 24 GB71 layers on system RAM~5.7 tok/s with offloadOpen →
AMD Radeon™ RX 7900 XTX181.2 GB of 24 GB71 layers on system RAM~5.7 tok/s with offloadOpen →
NVIDIA GeForce RTX 3090181.2 GB of 24 GB71 layers on system RAM~5.7 tok/s with offloadOpen →
NVIDIA GeForce RTX 5090181.2 GB of 32 GB68 layers on system RAM~6.0 tok/s with offloadOpen →
NVIDIA RTX 6000 Ada Generation181.2 GB of 48 GB60 layers on system RAM~6.6 tok/s with offloadOpen →
Apple M4 Pro181.2 GB of 64 GBdoes not fitnot calibratedOpen →
Apple M5 Pro181.2 GB of 64 GBdoes not fitnot calibratedOpen →
AMD Ryzen AI Max+ 395 with Radeon 8060S181.2 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M3 Max181.2 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M4 Max181.2 GB of 128 GBdoes not fitnot calibratedOpen →
NVIDIA DGX Spark181.2 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M2 Ultra181.2 GB of 192 GBfits~41 tok/sOpen →

Decode is a calibrated estimate from the published calibration; capacity is the manufacturer's published ceiling, not guaranteed free memory. Fitting in memory is not the same as loading: whether the runtime and version you have supports this architecture and format on that machine has not been tested here. Offloaded rows assume there is enough system RAM for the overflow — the calculator checks that against the RAM you declare. Each “Open” link keeps this model, format and context.

Run it in the cloud

Rent a GPU that runs Hy3-preview

Every rentable machine that holds the whole model, from Vast.ai and RunPod, with this site's speed estimate for each and the provider's own price, read live. Cheapest first; sort by cost per token or speed instead, or switch the format.

Prices · 18:30 UTC, 22 Sept
10 configurations in stock · 10 out of stock
MachineSpeedPer hourPer M tokensWhereRent
4× L40S192 GBcheapest~166 tok/sFaster than you read$1.87$3.12Vast.aiMarketplace · 99.2% reliableRent on Vast.ai
2× RTX PRO 6000192 GB~172 tok/sFaster than you read$2.27$3.65Vast.aiMarketplace · 98.8% reliableRent on Vast.ai
4× RTX 6000 Ada192 GB~184 tok/sFaster than you read$3.00$4.51Vast.aiMarketplace · 98.5% reliableRent on Vast.ai
4× RTX PRO 5000192 GB~258 tok/sFaster than you read$3.28$3.52RunPodCommunity CloudRent on RunPod
8× RTX 4090192 GBrecommendedbest value~388 tok/sFaster than you read$4.06$2.90Vast.aiMarketplace · 99.2% reliableRent on Vast.ai
2× H200 NVL282 GB~462 tok/sFaster than you read$7.20$4.33Vast.aiMarketplace · 98.7% reliableRent on Vast.ai
1× B300270 GB~370 tok/sFaster than you read$7.89$5.91RunPodSecure CloudRent on RunPod
4× H100 SXM320 GBfastest~645 tok/sFaster than you read$9.07$3.90Vast.aiMarketplace · 99.9% reliableRunPod $13.96/h Rent on Vast.ai

No GPU at all

Use Hy3-preview by the token

For one person chatting, that is about 4.8× cheaper than the best-value rented card above ($2.90 per million tokens). A rented GPU pays off when you keep it busy — many requests batched together, long agent runs — or when the data must stay on a machine you control, or the exact file you want is not served anywhere.

First time renting a GPU? How it works, in four steps
  1. Create an account and add credit. Both providers are prepaid: $10 is enough to try any card on this page for hours. A new RunPod account that arrives through a referral link gets a one-time $5 credit when it first adds $10.
  2. Pick the card and the count shown here, and a template: llama.cpp or vLLM for an OpenAI-compatible API, or one with a web chat if you only want to talk to the model. “How to launch it” below gives the exact image and command.
  3. Wait for the download. The weights are fetched on the machine; tens of gigabytes take a few minutes on a data-centre connection. You pay by the second from the moment it starts.
  4. Stop it when you are done. A running machine bills even when idle. A stopped one costs nothing per hour, though its disk is usually still billed until you delete it.
How to launch it

No Q4_K_M file of Hy3-preview is catalogued here, so there is no exact command to vouch for. Start the machine from the llama.cpp server image ghcr.io/ggml-org/llama.cpp:server-cuda and point -hf at a Q4_K_M build from the publisher or a quantizer you trust — search Hugging Face for one. Check its file size against the memory figure on this page before you rent.

Every machine holds the whole model at Q4_K_M and 8,192 tokens of context, no offload. Speeds are this site’s single-stream decode estimates; prices are what each provider’s own API quoted, on-demand, for the whole machine. Vast hosts below 98% measured reliability are left out.

Referral links Vast.ai, RunPod and Novita pay us a share of what you spend if you sign up through these buttons. It costs you nothing, and it never decides an order or a recommendation: both are computed from the live price and the speed, and options that pay us nothing are listed and recommended on the same terms. How we rank

Buying a card instead, or paying by the token? Compare a year of Hy3-preview three ways →

Under the hood

Architecture, read from the publisher’s file

The numbers every figure above is computed from, with the file they came from.

Architecture

✓ Architecture read from the published config.json

Retrieved 2026-09-01 at pinned commit 549c2b3a0fd5; parameter count from the safetensors index at the same revision.

Architecture
hy_v3
Layers
80
Hidden size
4,096
Attention heads
64
KV heads
8
Head dimension
128
Feed-forward width
13,312
Vocabulary
120,832
Context ceiling
262,144
RoPE theta
11,158,840
Experts
192
Experts / token
8
Expert width
1,536
Shared experts
1

Where the memory goes

tokenembedding× 80 decoder blocksGrouped-query attention64 query · 8 KV headsfull contextRouted experts8 of 192 per tokenall residentoutputprojectiongrows with contextcapacity ≠ traffic

Grouped-query attention shares each key/value head across 8 query heads, so the KV cache is 13% of what multi-head attention would need at the same context.

Open the exact official config.json ↗

More from Tencent

Hunyuan-A13B-InstructHy-MT2-1.8BHy-MT2-30B-A3BHy-MT2-7BHy3Hy4-preview

Use this answer in your own product

As JSON, from the API
curl -s https://llmbottleneck.com/v1/analyze \
  -H "Authorization: Bearer $LLMB_KEY" -H "content-type: application/json" \
  -d '{"model":"tencent-hy3-preview","quantization":"Q4_K_M","context":8192,"hardware":"nvidia-geforce-rtx-4090"}'

Same engine, same evidence, every field sourced. Free key in one step →

On your page, as a widget
<script src="https://llmbottleneck.com/widget.js"
  data-model="tencent-hy3-preview" data-quantization="Q4_K_M"
  data-hardware="nvidia-geforce-rtx-4090" data-context="8192"></script>

No key needed. Unbranded, with your own buy button, on Pro and Business →

Citing this page

LLM Bottleneck. “Hy3-preview VRAM and hardware requirements.” Architecture from tencent/Hy3-preview at revision 549c2b3a0fd5, retrieved 2026-09-01. https://llmbottleneck.com/models/tencent-hy3-preview

Every figure above is either the published value or a reconstruction whose measured error is on the accuracy page.

llmbottleneck
catalogue 2026-10-03models 327devices 135