Quancisuancis

Help

Not available yet

What Kael and the platform don't do today, stated plainly.

Kael is in beta. This is what it does not do today. If something here is missing for you, tell us at support@quancis.space.

Inputs and outputs

  • Files: PDFs, audio, video, and other file types. Images are supported. See Images.
  • Output other than text: Kael does not create images, audio, or files.
  • Languages other than English. Kael is built and tested for English. Other languages have not been tested, so they are not listed as supported.

Controls

  • Auto thinking level: there is no auto. Choose off, low, high, max, z-low or z-high yourself. See Thinking levels.
  • Sampling: temperature, top_p, seed and similar settings are ignored. Output is not deterministic.
  • Response length: max_tokens is ignored. Ask for the length you want in your prompt.
  • Structured output: response_format is not enforced, and tool_choice is not enforced. Ask for JSON in your prompt and validate it. See Request fields.
  • Built-in tools: web search, code execution, and file search are not available. Your own function tools are.
  • Answer caching: answers are never cached. Input caching exists and is automatic.

Platform

  • Key permissions and spend caps: a key can use every endpoint and your whole balance.
  • Other APIs: no embeddings, batch API, fine-tuning, or file uploads.
  • Downloadable or self-hosted weights: Kael is available only through the API and Quancis apps.

Not published yet

  • A model card or technical report.
  • An uptime commitment or SLA, and worst-case latency figures for long prompts.
  • Benchmark results. Quancis does not publish scores it ran itself.

Versioning

Minor bug fixes ship silently under the same model ID. A major change gets a new model ID, such as kael-beta-1-1, so you choose when to move. Details can change during the beta as we learn from real use.