AI Inference API
Open-source AI models behind an OpenAI-compatible API, hosted in Bulgaria. Pay per token from prepaid credit.
Open models, no servers to run
You send a request and get an answer - we take care of the graphics cards, the model weights and the scaling. Language models for chat and reasoning, embeddings for search and Whisper for speech to text are ready to use, and we can host any other open-source model on request.
Works with the code you already have
The API follows the OpenAI format. In most applications and libraries you change only the base URL and the API key, and tools such as n8n, LangChain and Open WebUI work without changes.
Your data stays in Bulgaria
Requests are processed on hardware hosted in Bulgaria, inside the European Union, instead of being sent to a provider overseas. That makes it far easier to use AI with customer data and internal documents under GDPR.
Key benefits
OpenAI-compatible
The same request format and SDKs you already use, with a different base URL.
Pay per token
No monthly fee and no idle servers. You pay only for the tokens you process.
Prepaid credit
Top up in euro and spend it as you go, with no surprise invoice.
Hosted in Bulgaria
Models run on hardware in Bulgaria, inside the European Union.
Any open-source model
Need a model that is not on the list? We host it for you on request.
Chat, embeddings and speech
Language models, bge-m3 embeddings for search and RAG, and Whisper for transcription in one API.
Model prices
Open-source models served from Bulgaria, priced in euro. Pick a model and message us to get an API key.
| Model | Type | Input | Output | Unit | |
|---|---|---|---|---|---|
| gpt-oss-120b | Chat | 0.15 € | 0.60 € | per 1M tokens | Inquire |
| gpt-oss-20b | Chat | 0.10 € | 0.20 € | per 1M tokens | Inquire |
| Qwen3.8-27B | Chat, reasoning | 0.40 € | 2.70 € | per 1M tokens | Inquire |
| Qwen3.5-397B | Chat, reasoning | 0.60 € | 3.60 € | per 1M tokens | Inquire |
| Qwen3.5-9B | Chat | 0.10 € | 0.15 € | per 1M tokens | Inquire |
| Qwen3-Coder-Next 80B | Code | 0.15 € | 0.80 € | per 1M tokens | Inquire |
| DeepSeek V4.1 Flash | Chat, reasoning | 0.60 € | 1.80 € | per 1M tokens | Inquire |
| GLM-5.3 | Chat, reasoning | 1.80 € | 5.50 € | per 1M tokens | Inquire |
| GLM-5.3 Flash | Chat | 0.15 € | 0.50 € | per 1M tokens | Inquire |
| Kimi K2.6 | Chat, reasoning | 0.60 € | 3.00 € | per 1M tokens | Inquire |
| Gemma 4 31B | Chat | 0.25 € | 0.50 € | per 1M tokens | Inquire |
| Mistral Small 4 | Chat | 0.20 € | 0.70 € | per 1M tokens | Inquire |
| Mistral Small 3.2 24B | Chat | 0.10 € | 0.30 € | per 1M tokens | Inquire |
| bge-m3 | Embeddings | 0.01 € | - | per 1M tokens | Inquire |
| Qwen3-Embedding-8B | Embeddings | 0.10 € | - | per 1M tokens | Inquire |
| Whisper large v3 | Speech to text | 0.0025 € | - | per audio minute | Inquire |
| Whisper large v3 turbo | Speech to text | 0.001 € | - | per audio minute | Inquire |
Usage is billed from prepaid credit. Need a model that is not listed? Ask us - we can host any open-source model.
Frequently asked questions
Pick a model in the price table and send the inquiry. We set up your account, load your credit and send you an API key.
Chat and reasoning: gpt-oss-120b and 20b, Qwen3.8-27B, Qwen3.5-397B and 9B, DeepSeek V4.1 Flash, GLM-5.3 and GLM-5.3 Flash, Kimi K2.6, Gemma 4 31B, Mistral Small 4 and 3.2. Coding: Qwen3-Coder-Next. Embeddings: bge-m3 and Qwen3-Embedding-8B. Speech to text: Whisper large v3 and turbo. Other open-source models are added on request.
You load prepaid credit in euro. Every request deducts the tokens it used at the price of the model, with input and output tokens priced separately.
Yes. The API follows the OpenAI format, so the official SDKs and most AI tools work by changing the base URL and the API key.
On hardware hosted in Bulgaria, inside the European Union.
Yes. For steady high volume or a private model we can dedicate a GPU to you at an hourly price - see our GPU Cloud page or ask us for an offer.

