LLMBOTTLENECK.COM
Tencent / hunyuan_v1_moe

Hunyuan-A13B-Instruct

80.4 billion parameters.

Open in the calculator →

Architecturepublished data
Quick answer
Hunyuan-A13B-Instruct needs about 50.7 GB at Q4_K_M — the format most people download — with 8,192 tokens of context. It fits entirely on 7 of 18 common devices, starting with the Apple M4 Pro (64 GB). What else fits in 64 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 M4 Pro, 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
FP16159.4 GB157.3 GB – 161.6 GBreconstructed size
Q8_085.4 GBexactpublished data
Q6_K65.4 GB64.5 GB – 66.3 GBreconstructed size
Q5_K_M56.6 GB55.9 GB – 57.4 GBreconstructed size
Q5_054.9 GB54.2 GB – 55.6 GBreconstructed size
Q4_K_M48.8 GBexactpublished data
Q4_045.4 GBexactpublished data
Q3_K_M39.6 GB36.4 GB – 42.7 GBreconstructed size
Q2_K29.1 GB27.7 GB – 30.5 GBreconstructed size

3 of these are published files. Open the pinned GGUF repository ↗

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 · 7 of 18 devices hold it entirely
DeviceMemoryVerdictDecodeCalculator
NVIDIA GeForce RTX 507050.7 GB of 12 GB26 layers on system RAM~1.5 tok/s with offloadOpen →
AMD Radeon™ RX 9070 XT50.7 GB of 16 GB23 layers on system RAM~1.6 tok/s with offloadOpen →
NVIDIA GeForce RTX 408050.7 GB of 16 GB23 layers on system RAM~1.6 tok/s with offloadOpen →
NVIDIA GeForce RTX 508050.7 GB of 16 GB23 layers on system RAM~1.6 tok/s with offloadOpen →
NVIDIA GeForce RTX 5070 Ti50.7 GB of 16 GB23 layers on system RAM~1.6 tok/s with offloadOpen →
NVIDIA GeForce RTX 5060 Ti 16GB50.7 GB of 16 GB23 layers on system RAM~1.6 tok/s with offloadOpen →
NVIDIA GeForce RTX 409050.7 GB of 24 GB18 layers on system RAM~2.0 tok/s with offloadOpen →
AMD Radeon™ RX 7900 XTX50.7 GB of 24 GB18 layers on system RAM~2.1 tok/s with offloadOpen →
NVIDIA GeForce RTX 309050.7 GB of 24 GB18 layers on system RAM~2.0 tok/s with offloadOpen →
NVIDIA GeForce RTX 509050.7 GB of 32 GB13 layers on system RAM~2.8 tok/s with offloadOpen →
NVIDIA RTX 6000 Ada Generation50.7 GB of 48 GB2 layers on system RAM~8.2 tok/s with offloadOpen →
Apple M4 Pro50.7 GB of 64 GBfits~5.0 tok/sSlowOpen →
Apple M5 Pro50.7 GB of 64 GBfits~5.6 tok/sSlowOpen →
AMD Ryzen AI Max+ 395 with Radeon 8060S50.7 GB of 128 GBfits~4.0 tok/sSlowOpen →
Apple M3 Max50.7 GB of 128 GBfits~7.2 tok/sSlowOpen →
Apple M4 Max50.7 GB of 128 GBfits~9.7 tok/sSlowOpen →
NVIDIA DGX Spark50.7 GB of 128 GBfits~3.7 tok/sSlowOpen →
Apple M2 Ultra50.7 GB of 192 GBfits~14 tok/sAbout reading paceOpen →

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 Hunyuan-A13B-Instruct

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
19 configurations in stock · 14 out of stock
MachineSpeedPer hourPer M tokensWhereRent
4× RTX 5060 Ti 16GB64 GBcheapest~24 tok/sAbout reading pace$0.51$5.79Vast.aiMarketplace · 99.5% reliableRent on Vast.ai
2× RTX 509064 GBrecommendedbest value~48 tok/sFaster than you read$0.54$3.05Vast.aiMarketplace · 99.6% reliableRent on Vast.ai
8× RTX 306096 GB~39 tok/sFaster than you read$0.59$4.17Vast.aiMarketplace · 99.7% reliableRent on Vast.ai
4× RTX 309096 GB~51 tok/sFaster than you read$0.60$3.25Vast.aiMarketplace · 99.4% reliableRent on Vast.ai
4× RTX 4060 Ti 16GB64 GB~15 tok/sAbout reading pace$0.80$14.19Vast.aiMarketplace · 99.3% reliableRent on Vast.ai
2× RTX 6000 Ada96 GB~26 tok/sAbout reading pace$1.07$11.37Vast.aiMarketplace · 99.4% reliableRent on Vast.ai
1× RTX PRO 600096 GB~24 tok/sAbout reading pace$1.14$12.93Vast.aiMarketplace · 98.8% reliableRunPod $1.69/h Rent on Vast.ai
1× A100 80GB PCIe80 GB~26 tok/sAbout reading pace$1.19$12.52RunPodCommunity CloudRent on RunPod

No GPU at all

Use Hunyuan-A13B-Instruct by the token

For one person chatting, that is about 5.3× cheaper than the best-value rented card above ($3.05 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 Q4_K_M on it (llama.cpp)
Container image
ghcr.io/ggml-org/llama.cpp:server-cuda
Start command / arguments
-hf tencent/Hunyuan-A13B-Instruct-GGUF:Q4_K_M -c 8192 -ngl 999 --host 0.0.0.0 --port 8080
Expose port
8080 — an OpenAI-compatible API at /v1

Starting from a saved llama.cpp template instead? Leave the arguments empty and set these variables — only the first two change between models:

LLAMA_ARG_HF_REPO=tencent/Hunyuan-A13B-Instruct-GGUF:Q4_K_M
LLAMA_ARG_CTX_SIZE=8192
LLAMA_ARG_N_GPU_LAYERS=999
LLAMA_ARG_HOST=0.0.0.0
LLAMA_ARG_PORT=8080

Already have the runtime? llama-server -hf tencent/Hunyuan-A13B-Instruct-GGUF:Q4_K_M -c 8192 -ngl 999 --host 0.0.0.0 --port 8080. Weights: tencent/Hunyuan-A13B-Instruct-GGUF. The CUDA image is for NVIDIA cards; an AMD machine needs llama.cpp's ROCm build.

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 Hunyuan-A13B-Instruct 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 290ddb9a56ed; parameter count from the safetensors index at the same revision.

Architecture
hunyuan_v1_moe
Layers
32
Hidden size
4,096
Attention heads
32
KV heads
8
Head dimension
128
Feed-forward width
3,072
Vocabulary
128,167
Context ceiling
32,768
RoPE theta
10,000
Experts
64
Expert width
3,072

Where the memory goes

tokenembedding× 32 decoder blocksGrouped-query attention32 query · 8 KV headsfull contextRouted expertsone networkall residentoutputprojectiongrows with contextcapacity ≠ traffic

Grouped-query attention shares each key/value head across 4 query heads, so the KV cache is 25% of what multi-head attention would need at the same context.

Open the exact official config.json ↗

Hunyuan-A13B-Instruct, card by card

One page per device: whether Hunyuan-A13B-Instruct fits, at which formats, and how fast.

More from Tencent

Hy-MT2-1.8BHy-MT2-30B-A3BHy-MT2-7BHy3Hy3-previewHy4-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-hunyuan-a13b-instruct","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-hunyuan-a13b-instruct" 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. “Hunyuan-A13B-Instruct VRAM and hardware requirements.” Architecture from tencent/Hunyuan-A13B-Instruct at revision 290ddb9a56ed, retrieved 2026-09-01. https://llmbottleneck.com/models/tencent-hunyuan-a13b-instruct

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