What models can do
The model catalogue is served at run time and differs by Account. This page carries the model facts that live in the code.
本文目前仅提供英文版本。阅读英文版。
- AI Studio
- 网页版
The list of models you can pick is not held in the product's code. It is served to the browser when a thread opens, and it differs by Account, so no page can name the models you will see. Open the model picker to read your own list: see Choose a model.
What this page carries instead is what the interface itself knows about models, whatever the catalogue says.
What you can attach to a message
A model accepts files by family. The picker offers a file only when the catalogue says the model takes that family. Generated from evolve-front-end/app/[lng]/studio/utils/model-file-support.ts and evolve-front-end/app/[lng]/studio/utils/supported-file-types.ts by evolve-front-end/scripts/help-generate.mjs.
| Family | Extensions |
|---|---|
| Images | .gif, .jpeg, .jpg, .png, .webp |
| Video | .mp4, .webm |
| Documents and code | .bash, .c, .cfg, .conf, .cpp, .cs, .css, .csv, .doc, .docx, .env, .go, .graphql, .html, .ini, .java, .js, .json, .jsx, .kt, .latex, .less, .log, .md, .pdf, .php, .ppt, .pptx, .proto, .py, .r, .rb, .rs, .rtf, .scala, .scss, .sh, .sql, .swift, .tex, .toml, .ts, .tsv, .tsx, .txt, .xls, .xlsx, .xml, .yaml, .yml |
Images are sent as image/gif, image/jpeg, image/png, image/webp, and video as video/mp4, video/webm. A document's MIME type follows its extension.
Reasoning levels
A model that exposes Reasoning offers some of 8 effort levels. The interface ranks them from least to most effort in this order: none, minimal, low, medium, high, xhigh, max, ultra. Which of them a model offers comes from the catalogue, and a model that offers none shows This model doesn't support any customizations. Generated from evolve-front-end/app/[lng]/studio/components/Chat/lower-effort-retry.ts by evolve-front-end/scripts/help-generate.mjs.
A level the catalogue names but this order does not carry, such as a provider's own wording, is shown as it arrives and is not ranked.
Image generation
Thread Image mode selects gpt-image-2.5-flare and returns pictures in the current thread. The standalone Media composer defaults to gpt-image-2. See Generate images and videos.
When the catalogue is silent
The catalogue reports what each model accepts and whether it can read the web. When a model's record says neither, the interface falls back to two lists of model names it holds itself: 35 names treated as able to read images, and 4 treated as able to read the web. Both lists are marked in the code as superseded by the catalogue. Generated from evolve-front-end/app/[lng]/studio/utils/modelUtils.ts by evolve-front-end/scripts/help-generate.mjs.
The names in those lists are not a list of models you can pick, so they are not reproduced here. Your own list is the model picker.
What this page cannot tell you
- Which models your Account allows. The catalogue is per Account and arrives at run time.
- What a model costs. See What your requests cost.
- How large a thread a model allows. See Token limits in a thread.
Related
相关页面
- Choose a model — Select the model name at the top of a thread, pick a group, then pick a model. A model picked from a provider group becomes your default.
- Token limits in a thread — Every thread has a token budget set by its model. The counter beside the message box shows how much is used, and the thread stops at the limit.
- No models are available for your account — When AI Studio offers no model, or refuses the one you chose, the cause is your Account's model policy, your plan, or the model's own limits.