LLMBOTTLENECK.COM
thinkingmachines / inkling_mm_model

Inkling-Small

266.0 billion parameters, routing 6 of 256 experts per token.

Open in the calculator →

Architecturepublished data
Quick answer
Inkling-Small needs about 165.5 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
FP16532.1 GB372.3 GB – 537.2 GBsize range
Q8_0282.8 GB197.8 GB – 287.0 GBsize range
Q6_K218.4 GB152.7 GB – 222.3 GBsize range
Q5_K_M190.0 GB128.0 GB – 222.3 GBsize range
Q5_0185.6 GB128.0 GB – 222.3 GBsize range
Q4_K_M163.3 GB104.7 GB – 222.3 GBsize range
Q4_0154.8 GB104.7 GB – 222.3 GBsize range
Q3_K_M133.0 GB80.0 GB – 222.3 GBsize range
Q2_K105.3 GB61.1 GB – 222.3 GBsize range

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 5070165.5 GB of 12 GB40 layers on system RAM~11 tok/s with offloadOpen →
AMD Radeon™ RX 9070 XT165.5 GB of 16 GB39 layers on system RAM~11 tok/s with offloadOpen →
NVIDIA GeForce RTX 4080165.5 GB of 16 GB39 layers on system RAM~11 tok/s with offloadOpen →
NVIDIA GeForce RTX 5080165.5 GB of 16 GB39 layers on system RAM~11 tok/s with offloadOpen →
NVIDIA GeForce RTX 5070 Ti165.5 GB of 16 GB39 layers on system RAM~11 tok/s with offloadOpen →
NVIDIA GeForce RTX 5060 Ti 16GB165.5 GB of 16 GB39 layers on system RAM~11 tok/s with offloadOpen →
NVIDIA GeForce RTX 4090165.5 GB of 24 GB37 layers on system RAM~12 tok/s with offloadOpen →
AMD Radeon™ RX 7900 XTX165.5 GB of 24 GB37 layers on system RAM~12 tok/s with offloadOpen →
NVIDIA GeForce RTX 3090165.5 GB of 24 GB37 layers on system RAM~11 tok/s with offloadOpen →
NVIDIA GeForce RTX 5090165.5 GB of 32 GB35 layers on system RAM~12 tok/s with offloadOpen →
NVIDIA RTX 6000 Ada Generation165.5 GB of 48 GB31 layers on system RAM~13 tok/s with offloadOpen →
Apple M4 Pro165.5 GB of 64 GBdoes not fitnot calibratedOpen →
Apple M5 Pro165.5 GB of 64 GBdoes not fitnot calibratedOpen →
AMD Ryzen AI Max+ 395 with Radeon 8060S165.5 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M3 Max165.5 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M4 Max165.5 GB of 128 GBdoes not fitnot calibratedOpen →
NVIDIA DGX Spark165.5 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M2 Ultra165.5 GB of 192 GBfits~66 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 Inkling-Small

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~329 tok/sFaster than you read$1.87$1.57Vast.aiMarketplace · 99.2% reliableRent on Vast.ai
2× RTX PRO 6000192 GB~342 tok/sFaster than you read$2.27$1.84Vast.aiMarketplace · 98.8% reliableRent on Vast.ai
4× RTX 6000 Ada192 GB~366 tok/sFaster than you read$3.00$2.27Vast.aiMarketplace · 98.5% reliableRent on Vast.ai
4× RTX PRO 5000192 GB~513 tok/sFaster than you read$3.28$1.78RunPodCommunity CloudRent on RunPod
8× RTX 4090192 GBrecommendedbest value~769 tok/sFaster than you read$4.06$1.46Vast.aiMarketplace · 99.2% reliableRent on Vast.ai
2× H200 NVL282 GB~916 tok/sFaster than you read$7.20$2.18Vast.aiMarketplace · 98.7% reliableRent on Vast.ai
1× B300270 GB~735 tok/sFaster than you read$7.89$2.98RunPodSecure CloudRent on RunPod
4× H100 SXM320 GBfastest~1279 tok/sFaster than you read$9.07$1.97Vast.aiMarketplace · 99.9% reliableRunPod $13.96/h Rent on Vast.ai

No GPU at all

Use Inkling-Small by the token

For one person chatting, that is about 1.2× cheaper than the best-value rented card above ($1.46 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 Inkling-Small 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 Inkling-Small 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 8cc5877b44d3; parameter count from the safetensors index at the same revision.

Architecture
inkling_mm_model
Layers
42
Hidden size
4,096
Attention heads
32
KV heads
8
Head dimension
128
Feed-forward width
2,048
Vocabulary
201,024
Experts
256
Experts / token
6
Shared experts
2

Where the memory goes

tokenembedding× 42 decoder blocksGrouped-query attention32 query · 8 KV headsfull contextRouted experts6 of 256 per tokenall 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.

This is a multimodal checkpoint (vision + audio); the diagram and memory sizing above cover the text decoder. Encoder/audio/image activations are not included in the KV or weight figures.

Open the exact official config.json ↗

More from thinkingmachines

Inkling

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":"thinkingmachines-inkling-small","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="thinkingmachines-inkling-small" 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. “Inkling-Small VRAM and hardware requirements.” Architecture from thinkingmachines/Inkling-Small at revision 8cc5877b44d3, retrieved 2026-09-01. https://llmbottleneck.com/models/thinkingmachines-inkling-small

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