deepseek-v4-flash missing image modality in model definition

[LifetimeVip]

Description

The model deepseek-v4-flash supports image input (multimodal), but the model definition in OpenCode does not have image listed in its input modalities.

Impact

When a user sends an image to OpenCode while using deepseek-v4-flash, the following error is returned:

ERROR: Cannot read "image.png" (this model does not support image input). Inform the user.

Expected Behavior

deepseek-v4-flash should have image: true in its capabilities.input so that OpenCode allows sending images when using this model.

Workaround

Users can add a modality override in their opencode.jsonc config:

{
  "provider": {
    "deepseek": {
      "models": {
        "deepseek-v4-flash": {
          "modalities": {
            "input": ["text", "image"],
            "output": ["text"]
          }
        }
      }
    }
  }
}

This should be added to the upstream model definition in models.dev or the snapshot, so it works out of the box.

[rekram1-node]

[rekram1-node]

maybe we arent sending images correctly forr deepseek but if i enable it it fails

are u sure it accepts image inputs? im not sure

[rekram1-node]

Yeah openrouter says the same thing it doesnt support images