AI compute, held onshore.
Shoreheld serves open-weight models from the UK, the EU, Switzerland and Singapore. You choose the region each request is processed in, and the model behind each id stays fixed.
Choose where a request runs
Processed in the United Kingdom.
Kimi K3.1 Pro: $2.20 per 1M input tokens, $11.00 per 1M output tokens.
curl \
https://uk.api.shoreheld.com/v1/chat/completions \
-H "Authorization: Bearer $SHOREHELD_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "kimi-k3.1-pro",
"messages": [{"role": "user", "content": "Summarise this contract clause."}]
}'How you can buy
Start with tokens. Move to dedicated or reserved capacity when your volume makes it cheaper.
| Way to buy | What it is | Unit | Price | Status |
|---|---|---|---|---|
| Serverless API | Pay for the tokens you use, with no minimum. | per 1M tokens | from $0.30 in, $1.20 out | Available now |
| Batch | Send work that can wait and get results within 24 hours. | per 1M tokens | half the serverless price | On request |
| Dedicated endpoint | One model on GPUs reserved for you, in one region. | per GPU-hour | from $1.85 | On request |
| Reserved node | Hardware dedicated to you in the country you choose, on a 12 to 36 month contract. | per reserved GPU-hour | from $1.70 | On request |
| Gateway and pilot | A gateway in your own cloud account, and a four-week test on your own code. | software free | pilot quoted | On request |
Models at launch
Open-weight models for chat and coding, with embedding and transcription models alongside.
| Model | Input per 1M tokens | Output per 1M tokens |
|---|---|---|
| DeepSeek V4.1 Flash | $0.30 | $1.20 |
| Kimi K3.1 Pro | $2.20 | $11.00 |
Also served: Qwen, Hy4, embedding models, and voice and transcription models. Ask us for ids and prices.
All models and idsWhat you get, and where the limits are
Each commitment is listed beside its limit.
| What you get | The limit |
|---|---|
| Each request is processed in the region you choose: the UK, the EU, Switzerland or Singapore. | A region is a country or bloc, not a named building. |
| Each model id maps to one published model and version. It does not change behind the id. | Open-weight models trail the best closed models on some work. Test them on your own tasks before you move. |
| The API follows the OpenAI format, so existing SDKs work when you change the base URL and the key. | Features that run on one lab's own servers, such as hosted tools and connectors, are not provided. |
Send each task to the right model
The Shoreheld gateway runs in your own cloud account. It checks each prompt for secrets and customer data before it leaves, checks what comes back, and blocks risky commands.
It sends everyday work to open models on Shoreheld and leaves the hardest tasks with the lab you already use. Engineers keep the tools they have.
How the gateway worksWorks with what you already run
The API follows the OpenAI format. Change the base URL and the key, and existing code keeps working.
Quickstartimport os
from openai import OpenAI
client = OpenAI(
base_url="https://uk.api.shoreheld.com/v1",
api_key=os.environ["SHOREHELD_API_KEY"],
)Start with an API key
Create a key, pick a region and make a request. For dedicated or reserved capacity, tell us the country and the size you need.