Skip to main content
deepseek-v4.1-flash-unrestricted supports image understanding alongside text. Use it to ask about a screenshot, diagram, photograph, or other supported image. glm-5.3-unrestricted accepts text only.

Chat Completions

Send an OpenAI-style image_url content block using either a base64 data URL or a public HTTPS image URL. Keep accompanying text in a text block. Image-only user messages are also supported.
Send PNG, JPEG, WebP or GIF images using the selected endpoint’s image content format. Use a public HTTPS URL or a base64 data URL. Audio, video, file attachments and image generation are unavailable. Send only images you intend to share with the inference service.

Playground

Select an image-capable model, attach a supported image, review its preview, and send it with a question. The interface shows the model’s reasoning setting and keeps reasoning separate from the answer. Remove an attachment before sending if it is not relevant. Image controls are unavailable for text-only models. Attachments stay in the browser’s current conversation rather than being saved as reusable uploaded files. Continuing a conversation can resend prior image content; it contributes to that request’s input and limits.

Limits and billing

The Playground accepts up to 8 images, 8 MiB per image and 16 MiB total per request, including conversation history. Base64 increases request size; the full JSON body must fit within 32 MiB. Consult the live catalog for current upload limits. Images and text share the selected model’s input/output context. Final billing uses reported input, cached-input and output tokens at that model’s rates. See models, limits, and billing.