LLMBOTTLENECK.COM
moonshotai / kimi_linear

Kimi-Linear-48B-A3B-Instruct

49.1 billion parameters, routing 8 of 256 experts per token.

Open in the calculator →

Architecturepublished data
Quick answer
Kimi-Linear-48B-A3B-Instruct needs about 30.4 GB at Q4_K_M — the format most people download — with 8,192 tokens of context. It fits entirely on 9 of 18 common devices, starting with the NVIDIA GeForce RTX 5090 (32 GB). What else fits in 32 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 NVIDIA GeForce RTX 5090, 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
FP1697.5 GB96.2 GB – 98.8 GBreconstructed size
Q8_051.8 GB51.2 GB – 52.5 GBreconstructed size
Q6_K40.0 GB39.5 GB – 40.6 GBreconstructed size
Q5_K_M34.6 GB34.1 GB – 35.0 GBreconstructed size
Q5_033.6 GB33.2 GB – 34.0 GBreconstructed size
Q4_K_M29.5 GB29.1 GB – 29.8 GBreconstructed size
Q4_027.6 GB27.2 GB – 27.9 GBreconstructed size
Q3_K_M24.2 GB22.3 GB – 26.1 GBreconstructed size
Q2_K17.8 GB17.0 GB – 18.7 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 · 9 of 18 devices hold it entirely
DeviceMemoryVerdictDecodeCalculator
NVIDIA GeForce RTX 507030.4 GB of 12 GB17 layers on system RAM~50 tok/s with offloadOpen →
AMD Radeon™ RX 9070 XT30.4 GB of 16 GB14 layers on system RAM~59 tok/s with offloadOpen →
NVIDIA GeForce RTX 408030.4 GB of 16 GB14 layers on system RAM~58 tok/s with offloadOpen →
NVIDIA GeForce RTX 508030.4 GB of 16 GB14 layers on system RAM~60 tok/s with offloadOpen →
NVIDIA GeForce RTX 5070 Ti30.4 GB of 16 GB14 layers on system RAM~60 tok/s with offloadOpen →
NVIDIA GeForce RTX 5060 Ti 16GB30.4 GB of 16 GB14 layers on system RAM~55 tok/s with offloadOpen →
NVIDIA GeForce RTX 409030.4 GB of 24 GB6 layers on system RAM~115 tok/s with offloadOpen →
AMD Radeon™ RX 7900 XTX30.4 GB of 24 GB6 layers on system RAM~118 tok/s with offloadOpen →
NVIDIA GeForce RTX 309030.4 GB of 24 GB6 layers on system RAM~113 tok/s with offloadOpen →
NVIDIA GeForce RTX 509030.4 GB of 32 GBfits~652 tok/sOpen →
NVIDIA RTX 6000 Ada Generation30.4 GB of 48 GBfits~349 tok/sOpen →
Apple M4 Pro30.4 GB of 64 GBfits~77 tok/sOpen →
Apple M5 Pro30.4 GB of 64 GBfits~82 tok/sOpen →
AMD Ryzen AI Max+ 395 with Radeon 8060S30.4 GB of 128 GBfits~107 tok/sOpen →
Apple M3 Max30.4 GB of 128 GBfits~94 tok/sOpen →
Apple M4 Max30.4 GB of 128 GBfits~107 tok/sOpen →
NVIDIA DGX Spark30.4 GB of 128 GBfits~99 tok/sOpen →
Apple M2 Ultra30.4 GB of 192 GBfits~122 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 Kimi-Linear-48B-A3B-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
23 configurations in stock · 10 out of stock
MachineSpeedPer hourPer M tokensWhereRent
4× RTX 306048 GBcheapest~443 tok/sFaster than you read$0.21$0.13Vast.aiMarketplace · 99.2% reliableRent on Vast.ai
2× RTX 4060 Ti 16GB32 GB~197 tok/sFaster than you read$0.22$0.32Vast.aiMarketplace · 99.7% reliableRent on Vast.ai
2× RTX 309048 GBrecommendedbest value~642 tok/sFaster than you read$0.24$0.10Vast.aiMarketplace · 99.5% reliableRent on Vast.ai
2× RTX 5060 Ti 16GB32 GB~307 tok/sFaster than you read$0.26$0.23Vast.aiMarketplace · 99.5% reliableRent on Vast.ai
1× RTX 6000 Ada48 GB~349 tok/sFaster than you read$0.26$0.21Vast.aiMarketplace · 99.9% reliableRunPod $0.84/h Rent on Vast.ai
1× RTX 509032 GB~651 tok/sFaster than you read$0.27$0.11Vast.aiMarketplace · 99.6% reliableRunPod $0.99/h Rent on Vast.ai
1× RTX A600048 GB~279 tok/sFaster than you read$0.28$0.28Vast.aiMarketplace · 99.6% reliableRunPod $0.53/h Rent on Vast.ai
1× RTX PRO 500048 GB~488 tok/sFaster than you read$0.51$0.29Vast.aiMarketplace · 98.1% reliableRunPod $0.82/h Rent on Vast.ai

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 Kimi-Linear-48B-A3B-Instruct 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 Kimi-Linear-48B-A3B-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 2026-09-13 at pinned commit e1df551a4471; parameter count from the safetensors index at the same revision.

Architecture
kimi_linear
Layers
27
Hidden size
2,304
Attention heads
32
KV heads
32
Head dimension
72
Feed-forward width
9,216
Vocabulary
163,840
RoPE theta
10,000
Experts
256
Experts / token
8
Expert width
1,024
Shared experts
1
Latent KV rank
512

Where the memory goes

tokenembedding× 27 decoder blocksLatent attentioncompressed KV rankfull contextRouted experts8 of 256 per tokenall residentoutputprojectiongrows with contextcapacity ≠ traffic

Latent attention projects keys and values into a compressed rank before caching them, which is why its KV cache is a fraction of a comparable dense model.

Open the exact official config.json ↗

Kimi-Linear-48B-A3B-Instruct, card by card

One page per device: whether Kimi-Linear-48B-A3B-Instruct fits, at which formats, and how fast.

More from moonshotai

Kimi-K2.5Kimi-K2.6Kimi-K2.7-CodeKimi-K2-InstructKimi-K2-Instruct-0905Kimi-K2-ThinkingKimi-K3

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":"moonshotai-kimi-linear-48b-a3b-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="moonshotai-kimi-linear-48b-a3b-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. “Kimi-Linear-48B-A3B-Instruct VRAM and hardware requirements.” Architecture from moonshotai/Kimi-Linear-48B-A3B-Instruct at revision e1df551a4471, retrieved 2026-09-13. https://llmbottleneck.com/models/moonshotai-kimi-linear-48b-a3b-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