EU and US zero-retention inference, encrypted history. Same prices in Chat and the API. From €0.08 / 1M input. Pay with Crypton credits or crypto.
Pick EU or US inference per model. Default zero retention at both providers. We do not log prompt bodies. Optional encrypted Chat history you can delete.
Read the AI PolicyEU models run in the EU on renewable energy. US models are a separate ZDR lane. You pick the model. Crypton is a UK company; billing metadata stays under UK law.
Default ZDR at the inference provider. We never log prompt bodies. Chat history is optional encrypted storage you can delete.
Chat and the OpenAI-compatible API use the same EUR rate card. No Chat surcharge. Paid usage comes from your Crypton credits.
Chat and the API on one Crypton account.
Open Crypton Chat in the app. History is encrypted, or use Incognito and nothing is stored.
Open ChatChat and the API share one EUR rate card. We do not compete with US routers on price. The US lane is residency and these models, not a cheap router. We sell ZDR, crypto, and one Crypton account.
| Model | Region | Context | Input | Cached | Output |
|---|---|---|---|---|---|
qwen3.5-9b |
EU | 256K | €0.08 | — | €0.25 |
gpt-oss-20b |
EU | 128K | €0.10 | — | €0.29 |
brick-complexity-pro |
EU | 100K | €0.10 | — | €0.28 |
qwen3.8-27b |
EU | 120K | €0.34 | — | €1.30 |
mistral-small-4-119b |
EU | 256K | €0.34 | — | €1.30 |
apertus-70b |
EU | 30K | €0.28 | — | €1.30 |
gemma4-31b |
EU | 100K | €0.28 | — | €1.30 |
gpt-oss-120b |
EU | 128K | €0.64 | — | €2.56 |
qwen3.5-122b |
EU | 256K | €0.64 | — | €2.56 |
glm5.2 |
EU | 96K | €1.24 | — | €3.16 |
kimi-k3 |
US ZDR | 1M | €3.31 | €0.33 | €16.56 |
deepseek-v4.1-flash |
US ZDR | 1M | €0.66 | €0.01 | €2.65 |
glm-5.3 |
US ZDR | 1M | €1.55 | €0.29 | €4.86 |
unrestricted-vision |
US ZDR Unrestricted | 256K | €1.10 | €0.11 | €3.31 |
glm-5.3-unrestricted |
US ZDR Unrestricted | 1M | €3.31 | €0.33 | €5.52 |
glm-5.2-unrestricted |
US ZDR Unrestricted | 1M | €3.31 | €0.33 | €5.52 |
Guests get 8 messages/day and 8,000 tokens/day on qwen3.5-9b. With an account: 20 messages/day and 25,000 tokens/day, 10 RPM. Larger models are paid-only. After the free bucket, keep chatting from Crypton credits.
Point any OpenAI SDK at our base URL and authenticate with a Crypton AI key.
from openai import OpenAI
client = OpenAI(
base_url="https://crypton.sh/api/ai/v1",
api_key="sk-crypton-...",
)
client.chat.completions.create(
model="qwen3.5-9b",
messages=[{"role":"user","content":"Hello from the EU."}],
)
The EU provider discards the request after the reply (third-party ZDR certificate). Both US providers’ published policies are the same: no request or response bodies stored. Crypton does not log prompt bodies. Signed-in Chat without Incognito stores encrypted history until you delete it. Guest Chat is not saved.
Yes. Guest Chat uses the free model with a smaller daily cap. Ask Crypton on the site also works without an account. Sign in for the larger free bucket, saved chats, paid models, and the API.
No. Say that out loud: Llama-class 70B on Crypton is still many times a cheap US router. OpenRouter and Groq mix cheaper US GPUs and routing; ZDR is not the default. We sell a certified EU ZDR lane and a US ZDR lane (published policy), one EUR rate card for Chat and the API, crypto/credits, and encrypted history you can delete. Buy us for residency and zero retention, not to beat a router on price.
No. Platform SMS filter and translation stay self-hosted. Support tickets and SimpleX are human. Ask Crypton answers product questions; it is not a support desk.
