# Inkling by Thinking Machines Lab: GPU Requirements and Cost
Source: https://savrn.com/models/inkling/gpus
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

## Every Accelerator, Every Precision

How many cards of each accelerator the SAVRN Index prices it takes to hold Inkling, and what that many cards cost an hour at the lowest listed on-demand price. Memory needed: 2.3 TB at 16-bit, 1.1 TB at 8-bit, 571 GB at 4-bit.

| Accelerator | Memory per card | Lowest price per card | 16-bit | 8-bit | 4-bit |
| --- | --- | --- | --- | --- | --- |
| [H100](https://savrn.com/ai-index/pricing/gpus/h100) Voltage Park | 80 GB | $1.99 | More than 8 | More than 8 | 8 cards $15.92/hr |
| [H200](https://savrn.com/ai-index/pricing/gpus/h200) GMI Cloud | 141 GB | $2.60 | More than 8 | More than 8 | 5 cards $13.00/hr |
| [B200](https://savrn.com/ai-index/pricing/gpus/b200) Vultr | 180 GB | $3.50 | More than 8 | 7 cards $24.50/hr | 4 cards $14.00/hr |
| [GB200 NVL72](https://savrn.com/ai-index/pricing/gpus/gb200) GMI Cloud | 186 GB | $8.00 | More than 8 | 7 cards $56.00/hr | 4 cards $32.00/hr |
| [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) Vultr | 192 GB | $1.85 | More than 8 | 6 cards $11.10/hr | 3 cards $5.55/hr |
| [MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) Vultr | 256 GB | $2.00 | More than 8 | 5 cards $10.00/hr | 3 cards $6.00/hr |
| [B300](https://savrn.com/ai-index/pricing/gpus/b300) Massed Compute | 268 GB | $6.60 | More than 8 | 5 cards $33.00/hr | 3 cards $19.80/hr |
| [GB300 NVL72](https://savrn.com/ai-index/pricing/gpus/gb300) Verda | 279 GB | $10.74 | More than 8 | 5 cards $53.70/hr | 3 cards $32.22/hr |
| [MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) Vultr | 288 GB | $2.59 | 8 cards $20.72/hr | 4 cards $10.36/hr | 2 cards $5.18/hr |

## Running It Around the Clock

| Precision | Cheapest setup | Per hour | Per month (730 hours) |
| --- | --- | --- | --- |
| 16-bit | 8x [MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) (Vultr) | $20.72 | $15,126 |
| 8-bit | 5x [MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) (Vultr) | $10.00 | $7,300 |
| 4-bit | 2x [MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) (Vultr) | $5.18 | $3,781 |

One copy of the model on rented cards, busy or idle. Serving more users at once takes more copies or more memory for their contexts.

## Questions

### How much VRAM does Inkling need?

About 2.3 TB at 16-bit; about 1.1 TB at 8-bit; about 571 GB at 4-bit: the weights plus 20% for the runtime and a short context. A long context needs more.

### What is the cheapest GPU setup to run Inkling?

At 16-bit, 8x MI355X from $20.72 an hour, at the lowest on-demand price the SAVRN Index lists.

### Can Inkling run on a single H100?

No. An H100 has 80 GB, and Inkling needs 571 GB even at 4-bit, so it takes more than one card.

Memory is the weights at that precision plus 20% for the runtime and a short context. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 12, 2026. Setups beyond eight cards, one server, are not listed.

## Inkling

- [Model card, license and download](https://savrn.com/models/inkling)
- [All image and text to text models](https://savrn.com/models/tasks/image-and-text-to-text)
- [GPU prices in the SAVRN Index](https://savrn.com/ai-index/pricing/gpus)