Models and pricing
One model today, more as the fleet grows. Prices in USD per million tokens, read live from the gateway document.
Qwen3.8 27B
FP8 weights on vLLM, on our own GPU hardware in Romania.| Model id | qwen/qwen3.8-27b |
|---|---|
| Endpoint | https://api.auxilio.ai/v1, OpenAI-compatible |
| Context | 128k tokens, prompt up to 124k |
| Max output | 32k tokens |
| Weights | FP8 on vLLM, own hardware |
| Input | $0.30 per million tokens |
| Cached input | $0.08 per million tokens |
| Output | $1.80 per million tokens, reasoning tokens billed as output |
| Capacity | 19.2M input and 800k output tokens per minute, 512 concurrent streams |
| Region | EU, datacenters in RO, own hardware |
| Data retention | None for prompts and completions. Policy |
| Status | Serving |
Features
- Streaming, with usage in every chunk
- Tool calls with automatic tool choice, JSON schema and structured outputs
- Thinking on by default, switched off per request (see the docs)
- Sampling controls: temperature, top_p, top_k, penalties, seed, stop
Limits
Over the per-model capacity the gateway answers 429 at once; back off and retry. One request up to 32 MB.
Billing
Per token used. Terms are agreed when your key is issued.
Live document
Every number on this page comes from api.auxilio.ai/openrouter/models, the same document the gateway publishes to marketplaces.
GPU servers
Need the whole card? The same fleet is available as virtual machines with one to eight GPUs. GPU Cloud