Models and prices
The models on the Ertas AI Gateway, their model IDs, which MedErtas model to choose, and the price per 1M input, cached and output tokens.
On this page
The Models page in the app lists every model you can call, with its model ID, context length and prices.
MedErtas
MedErtas is our family of open-weight medical models. Both are on the gateway in beta.
| Model ID | Model | Input | Cached input | Output |
|---|---|---|---|---|
ertas-ai/medertas-31b | MedErtas-v0.1-31B-LS | $0.20 | $0.05 | $0.60 |
ertas-ai/medertas-27b | MedErtas-v0.1-27B-LS | $0.10 | $0.05 | $1.00 |
Prices are in USD per 1M tokens, paid in credits at 100 credits per US$1.
Which one to use:
ertas-ai/medertas-27bfor clinical calculation and long patient records with short answers. Its input price is lower, which suits large inputs.ertas-ai/medertas-31bfor medical knowledge, medical codes and checking answers. Its output price is lower.
Evaluate accuracy, latency and total token use on your own workload before you choose. The weights for both are also on Hugging Face under Apache-2.0 if you'd rather run them yourself.
How tokens are billed
- Input tokens are everything you send: system prompt, messages, tool definitions.
- Cached input tokens are input the model has already processed in a recent request with the same opening, billed at the cached price.
- Output tokens are everything the model generates, including tool calls.
See Credits and usage for how charges reach your balance.
List models from code
curl https://gateway.ertas.ai/v1/models \
-H "Authorization: Bearer $ERTAS_API_KEY"