Model
Use ars-1 as the model ID in every request.
Limits
- id
- ars-1
- name
- Arscelare 1
- context
- 65,536 tokens
- max output
- 8,192 tokens
Messages are never truncated. The fully rendered prompt and requested output allowance must fit within the context window. If you omit an output limit, Arscelare reserves the full 8,192 tokens. If you send both output-limit fields, their values must match.
Reasoning
Reasoning is on by default at medium. Supported reasoning_effort values are none, low, medium, and xhigh. Use none to turn reasoning off. Other values, including high, return 400.
Effort values guide how thoroughly the model reasons. They are not fixed token budgets. Both max_tokens and max_completion_tokens limit reasoning and the visible answer together. Reaching that limit during reasoning can produce a billed response with no visible answer.
Reasoning appears in reasoning_content and is billed at the output rate, even if your client does not display it.
Sampling defaults
Omitted values use the defaults below. Values in your request take precedence.
| Parameter | Reasoning on | Reasoning off |
|---|---|---|
| temperature | 1.0 | 0.7 |
| top_p | 0.95 | 0.8 |
| top_k | 20 | 20 |
| min_p | 0 | 0 |
| presence_penalty | 0 | 1.5 |
| repeat_penalty | 1.0 | 1.0 |