Heabsy
Sovereign inference API for open models: EEA GPUs, zero retention
Last reviewed Sep 30, 2026
Heabsy is an OpenAI- and Anthropic-compatible inference API for open-weight models, operated by FEYA, s.r.o. in Bratislava, Slovakia, an EU company with no US parent. Models in the EEA tier, including Qwen3.8 27B, run on GPUs that Heabsy owns and operates inside the European Economic Area, with zero data retention: prompts and completions are processed in memory and never written to logs. A routed tier adds more than 25 open models (DeepSeek, Kimi, GLM, gpt-oss, Llama, Gemma, Qwen and others) bought from global providers under a European contract; every model in the catalog is labelled with where it is processed. One API key, one EU invoice, per-token pricing with no minimum spend or subscription.
LLM API Pricing
Pay-per-token pricing. Prices shown per 1M tokens.
Prices last updated: October 6, 2026
| Model | Input/1M | Output/1M | |||
|---|---|---|---|---|---|
| $0.040 | $0.120 | ||||
| $0.050 | $0.080 | ||||
| $0.050 | $0.200 | ||||
| $0.060 | $0.120 | ||||
| $0.070 | $0.340 | ||||
| $0.080 | $0.120 | ||||
| $0.090 | $0.620 | ||||
| $0.090 | $0.390 | ||||
| $0.110 | $0.510 | ||||
| $0.120 | $0.310 | ||||
Heabsy pricing by model
Input, output, and batch rates, plus alternatives, for one model at a time.
Pros & Cons
Advantages
- OpenAI- and Anthropic-compatible API: existing SDKs, Claude Code and Cursor switch by changing the base URL
- EEA tier runs on Heabsy's own GPUs inside the EEA with zero data retention; prompts are never logged
- EU company (Slovakia) with no US parent; DPA and sub-processor list published
- Every model is labelled with where it is processed: EEA only or outside the EEA
- Per-token pricing with no minimum spend or subscription; one EU invoice for all models
Limitations
- Most of the catalog is routed through third-party providers and may be processed outside the EEA; only a few models run on Heabsy's own hardware
- Accounts are opened on request; there is no free tier
- No embeddings endpoint and no GPU rental
- No ISO 27001 or SOC 2 certification
Key Features
OpenAI- and Anthropic-compatible API
Base URL https://api.heabsy.com/v1. Streaming, tool calling and structured output work as with the original SDKs; vision on supported models.
EEA tier with zero data retention
Models marked EEA only run on GPUs Heabsy operates inside the European Economic Area. Prompts and completions are processed in memory and never written to logs.
Routed catalog of open models
More than 25 open models from DeepSeek, Moonshot, Z.ai, OpenAI, Meta, Google, Qwen, NVIDIA and others, bought from global providers under a European contract and billed by Heabsy.
Long context and prefix caching
Up to 1M tokens of context on several models and 262K on Qwen3.8 27B; cached input is billed at a lower rate.
Per-model processing labels
The catalog at heabsy.com/models shows for each model whether it is processed only in the EEA or may leave it.
Pricing Options
| Option | Details |
|---|---|
| Pay per token | Per 1M tokens for input, cached input and output, listed per model at heabsy.com/models. No minimum spend, no subscription. |
| Dedicated deployments | Open models can be brought onto Heabsy's own EEA hardware when the workload justifies a dedicated fleet; priced on request. |
Availability & Support
Regions
EEA tier: Heabsy's own GPUs inside the European Economic Area, EEA-only processing. Routed tier: third-party providers under a European contract, may be processed outside the EEA. Each model is labelled in the catalog.
Support
Email hello@heabsy.com or the contact form at heabsy.com/contacts. API reference at api.heabsy.com/docs. Website in English, German, Slovak and Russian.
Company
Published by Heabsy to identify the company behind this listing.
- Legal entity
- FEYA, s.r.o.
- Headquarters
- 🇸🇰 Bratislava, SK
- Registration
- Slovak Business Register (IČO) 47 887 541
- Contact
- hello@heabsy.com
Getting Started
- 1
Request access
Write via heabsy.com/contacts; API keys are issued in the console at platform.heabsy.com.
- 2
Point your SDK at Heabsy
Set the base URL to https://api.heabsy.com/v1 in the OpenAI or Anthropic SDK and use your Heabsy key.
- 3
Choose a model
Pick a model ID from heabsy.com/models, for example qwen38 for the EEA tier.
Frequently Asked Questions
Which models does Heabsy serve?
Check the pricing table above for the models Heabsy currently serves and their per-token rates.
How do I get started with Heabsy?
Request access, Point your SDK at Heabsy, Choose a model
What are Heabsy's main advantages?
Heabsy's main advantages include: OpenAI- and Anthropic-compatible API: existing SDKs, Claude Code and Cursor switch by changing the base URL, EEA tier runs on Heabsy's own GPUs inside the EEA with zero data retention; prompts are never logged, EU company (Slovakia) with no US parent; DPA and sub-processor list published, Every model is labelled with where it is processed: EEA only or outside the EEA, Per-token pricing with no minimum spend or subscription; one EU invoice for all models.
What are Heabsy's limitations?
Heabsy's main limitations include: Most of the catalog is routed through third-party providers and may be processed outside the EEA; only a few models run on Heabsy's own hardware, Accounts are opened on request; there is no free tier, No embeddings endpoint and no GPU rental, No ISO 27001 or SOC 2 certification.
Compare Providers
Find the best prices for the same GPUs and models from other providers