Skip to main content
AskTable is not tied to any single LLM. On first start, the console’s setup wizard walks you through configuring a model step by step — you just need one model’s connection details. This page covers the background and how to get model compute.

​
Supported models

Anything reachable through one of these two protocols works:
  • OpenAI-compatible — OpenAI GPT, Alibaba Cloud Model Studio (Qwen), DeepSeek, and any OpenAI-compatible endpoint.
  • Anthropic — the Claude family.

​
Requirements

AskTable’s agents query data, run Python, and draw charts by calling tools, so a model must support function / tool calling. Use deepseek-v4-pro as the capability baseline — stronger models are more reliable, and a larger context window helps too.

​
Getting model compute

If you’d rather not bring your own model key, use AskTable’s hosted compute: apply for a model-compute API key on the Private deployment page in the AskTable cloud console and enter it in the wizard — no need to integrate each provider yourself. To use your own or an in-network model, choose Custom in the wizard and enter the model’s endpoint, key, and name.

​
Next steps