Products · Serverless Models
Choose a model. Call the API.
AxForge operates the infrastructure, scaling and serving layer. You get an API key and an OpenAI-compatible endpoint — no VM, no CUDA setup, no infrastructure management.
What you get
One key, every served model
| API | OpenAI-compatible — chat, embeddings, images, audio and music on api.axforge.ai/v1. Streaming, tool calls, usage reporting. |
|---|---|
| Models | Qwen3.8 27B, Qwen3 Embedding, ERNIE Image Turbo, FLUX.2 Klein edits, MiniMax Music 3, Whisper, Piper — the full catalogue. |
| Pricing | Usage-based, in EUR — from €0.29 / 1M input tokens. All prices. |
| Region | Pinned to an EU region on the key, echoed on every response. |
| Data | Zero prompt retention — prompts and completions are processed in memory and never stored. Trust centre. |
First request
Five minutes to a completion
curl https://api.axforge.ai/v1/chat/completions \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-d '{"model":"qwen3.8-27b-nvfp4",
"messages":[{"role":"user","content":"Hello"}]}'
Need more?
When serverless is not the shape
A model you want that is not on the shared API, traffic that needs its own hardware, or full control of the stack — that is Managed GPU or a GPU VM. Same hardware, same regions, different amount of control.