Free Tier · verified Nov 9, 2026
Hetzner: Free experimental OpenAI-compatible inference
Collected official Hetzner documentation reconfirms free OpenAI-compatible inference while experimental, with Qwen/Qwen3.6-35B-A3B-FP8 and Qwen3.8-27B documented at 262,144-token context with image and text input. Customers need Experiments access and an API token; availability is not guaranteed, production use is discouraged, and the authenticated /v1/models response is definitive for the rotating model list.
PrototypingMultimodalOpenAI-compatible
What you actually get
- Eligibility
- Hetzner customers with Experiments access and an inference API token; no startup or VC requirement stated
- Rate and usage limits
- Per API key: 3M input and 60K output tokens per 60 seconds; 500M input and 5M output tokens per 24 hours.
- Models
- Qwen/Qwen3.6-35B-A3B-FP8
- Card requirement
- unknown
- Regions
- Global
- After the free allowance
- Requests are rate-limited. Hetzner does not promise continued availability.
- Data and privacy catch
- Check Hetzner's current experiment terms before sending sensitive data.
Do not treat this as guaranteed capacity
The provider describes this access as experimental, rotating, or subject to change. Build a fallback before using it in anything important.