Returns the TextCortex models available through the OpenAI-compatible public API.

A model whose reasoning can be varied is listed once, under its base id; control its reasoning with reasoning_effort on the request rather than by choosing a model id. The reasoning-only minimax-m3-thinking model is listed under its own suffixed id, which is its name and not a variant — there is no base id for it. Deprecated -thinking and -non-thinking ids are no longer listed here, though they are still accepted as request input during the compatibility period.

This lists chat models. For image generation, use the model table on POST /images/generations.

Choose a chat model

Use the table below as a reference for the standard deployment. Your key's GET /models response determines which IDs you can actually use; tenant routes and policies can narrow controls or limits. The table is documentation, not extra fields in the model-list response. Image-generation models have a separate table on POST /images/generations.

Context includes input and output; leave room for both. A dash means no limit is advertised here. Output ceilings describe configured standard routes, not a promise that every request can use that many tokens. Function calling is exposed by these chat routes; your application executes the functions. response_format means the route supports the parameter, not every JSON Schema feature. Models without an advertised reasoning control can still reason; omission uses their default. Unsupported reasoning_effort can be ignored rather than rejected.

{
  "data": [
    {
      "id": "gpt-6-luna",
      "owned_by": "textcortex"
    }
  ]
}