🛠 Solheim.ai has launched 'Virtual Private LLM' — a dedicated LLM instance on servers in the EU with a fixed subscription and no token counter.
The service offers instances starting at €15/month for one instance with a 64k context. At launch, the Qwen3.6-35B-A3B (MoE architecture, 35B parameters, Apache 2.0) and DeepSeek V4 Flash (MIT, 256k context) models are available. The API is fully compatible with the OpenAI SDK — integration into Cline, Roo Code, and VS Code is done by replacing two environment variables. The GPU infrastructure is located in European data centers without the involvement of US hyperscalers.
🌍 Solheim solves the data sovereignty problem during inference — requests do not leave the EU jurisdiction, which meets GDPR and EU AI Act requirements without involving infrastructure subject to the US CLOUD Act. The 'rent a machine, not tokens' subscription model avoids the blocking of paid tools like Claude Code in the middle of work sessions due to a rolling usage window.
👤 If you need a predictable LLM backend for coding agents, automation, or prototypes without billing surprises — Solheim offers an OpenAI-compatible API starting at €15/month. The limited performance of open-weight models will be weaker than Claude and GPT on complex tasks, but for most dev tasks (diffs, tests, refactoring, tool-calling) it works quite well.
Source 1: https://solheim.ai Source 2: https://www.indiehackers.com/post/building-solheim-a-vps-for-llms-hosted-entirely-in-the-eu-edd536b19c