hixtube

Open-model inference pilot

Hixtube.

DeepSeek.
Room to think.

Abliterated DeepSeek V4.1 Flash 1017. Tested 1M context. ZDR inference in Finland. Ordinary TLS, not confidential or E2EE.

View on TrustedRouter

DEEPSEEK V4.1 FLASH / REVISION 1017

Compute illustration

One model.
A clear offering.

We serve dealignai's UNCENSORED FP8 checkpoint of DeepSeek V4.1 Flash 1017, not the original DeepSeek weights. Check current availability and pricing on TrustedRouter.

Model
DeepSeek V4.1 Flash
Revision
1017
Serving
Finland / ZDR
Access
TrustedRouter

Know what
you're using.

ZDR inference on a Verda GPU node in Finland, rented through Spheron. No persistent prompt or output storage and no training on your content. Ordinary HTTPS, not verified E2EE or confidential inference.

01 / VERIFY

TLS is not a secure enclave.

TLS protects network transport; it does not prevent the node operator from accessing workloads in memory. End-to-end hardware attestation is still being built and is not enabled for this route.

03 / EVALUATE

Start with your workload.

The 1M context test passed for one request at a time. High availability and sustained mixed-workload capacity are not yet qualified. Benchmark your workload before scaling.

DeepSeek inference
for your next workload.

Tell us what you're building, your expected volume and your security requirements. We'll discuss access and pricing.

Contact the team

Hixtube inquiries: help@trustedrouter.com