Supported models
The DIAL Vertex AI Adapter serves a fixed list of Vertex AI models, addressed by their Vertex AI model IDs in the adapter URL. This reference lists them with the input they take and the features they support. It is for platform engineers who choose which Vertex AI models to offer in DIAL; the list follows the adapter's development branch, so check the release you run.
Chat models
Every model below serves POST /openai/deployments/<model id>/chat/completions.
| Family | Models | Model IDs | Input → output |
|---|---|---|---|
| Gemini 3.5 | Flash | gemini-3.5-flash | Text, PDF, image, audio, video → text |
| Gemini 3.5 | Transcribe | gemini-3.5-transcribe-preview | Audio → text |
| Gemini 3.1 | Pro, Flash Lite, Flash Image | gemini-3.1-pro-preview, gemini-3.1-flash-lite[-preview], gemini-3.1-flash-image-preview | Text, PDF, image, audio, video → text |
| Gemini 3 | Pro, Flash | gemini-3-pro[-preview], gemini-3-flash-preview | Text, PDF, image, audio, video → text |
| Gemini 3 | Pro Image | gemini-3-pro-image-preview | Text, image → text, image |
| Gemini 2.5 | Flash, Pro | gemini-2.5-flash, gemini-2.5-pro | Text, PDF, image, audio, video → text |
| Gemini 2.5 | Flash Image | gemini-2.5-flash-image | Text, image → text, image |
| Gemini 2.0 | Flash, Flash Lite | gemini-2.0-flash-001, gemini-2.0-flash-lite-001, gemini-2.0-flash-exp | Text, PDF, image, audio, video → text; -exp also outputs images |
| Claude | 4.6 Opus and Sonnet | claude-opus-4-6, claude-sonnet-4-6 | PDF, text, image → text |
| Claude | 4.5 Opus, Sonnet, Haiku | claude-opus-4-5@20251101, claude-sonnet-4-5@20250929, claude-haiku-4-5@20251001 | PDF, text, image → text |
| Claude | 4.1 Opus, 4 Opus, 4 Sonnet | claude-opus-4-1@20250805, claude-opus-4@20250514, claude-sonnet-4@20250514 | PDF, text, image → text |
| Claude | 3.7 Sonnet, 3.5 Sonnet v2, 3.5 Sonnet, 3.5 Haiku | claude-3-7-sonnet@20250219, claude-3-5-sonnet-v2@20241022, claude-3-5-sonnet@20240620, claude-3-5-haiku@20241022 | PDF and text, most with images → text |
| Claude | 3 Opus, 3 Haiku | claude-3-opus@20240229, claude-3-haiku@20240307 | Text, image → text |
| Imagen | 4, 3, 2 | imagen-4.0-generate-001 and the other imagen-4.0-* and imagen-3.0-* IDs, imagegeneration@005 | Text → image |
| Veo | 3.1 Generate and Fast Generate | veo-3.1-generate-(001|preview), veo-3.1-fast-generate-(001|preview) | Text, image, video → video |
| Veo | 3.0 Generate and Fast Generate | veo-3.0-generate-(001|preview), veo-3.0-fast-generate-(001|preview) | Text, image → video |
| Mistral | Medium 3, Small 2503, Codestral 2 | mistral-medium-3, mistral-small-2503, codestral-2 | Text, most with images → text |
The full list of Imagen 4 IDs is imagen-4.0-generate-001, imagen-4.0-fast-generate-001, imagen-4.0-ultra-generate-001, and their -preview-06-06 variants; Imagen 3 is imagen-3.0-generate-001, imagen-3.0-generate-002, and imagen-3.0-fast-generate-001.
Features by family
| Family | /tokenize | /truncate_prompt and max_prompt_tokens | Tools | /configuration |
|---|---|---|---|---|
| Gemini 2.5 and later | Yes | Yes | Yes, except Gemini 3.5 Transcribe | Yes, except Gemini 3.5 Transcribe |
| Gemini 2.0 | Yes | Yes | Yes | No |
| Claude | Yes | Yes | Yes | Yes |
| Imagen, Veo | Yes | Yes | No | Yes |
| Mistral | No | No | Yes | No |
Models with special behavior
- Gemini 3.5 Transcribe accepts audio attachments only. It rejects system messages and text input, and the adapter sends only the last message, so each DIAL Chat message is transcribed on its own.
- Gemini 2.5 Flash Image and Gemini 3 Pro Image generate and edit images: a follow-up such as "replace the cat with a dog" edits the previous image. DIAL Chat removes attachments from assistant messages by default, so for editing in Chat, set
features.assistantAttachmentsInRequestSupported: trueon the model. - Veo takes the text and files of the last user message only. An image can come as an image content part or an attachment; a video, as an attachment.
Embedding models
These models serve POST /openai/deployments/<model id>/embeddings.
| Model | Model ID | Language | Input |
|---|---|---|---|
| Gemini Embedding 2 | gemini-embedding-2-preview | English | Text, image, video, audio, PDF |
| Gemini Embeddings | gemini-embedding-001 | Multilingual | Text |
| Multimodal embeddings | multimodalembedding@001 | English | Text, image |
| Embeddings for Text | text-embedding-004, text-embedding-005 | English | Text |
Multimodal input goes in custom_input as DIAL attachments, each with a MIME type and a url or base64 data. Several parts can form one composite input. Gemini Embedding 2 does not accept instruction prompts, so leave custom_fields.instruction unset.
A model not on the list
When Vertex AI releases a model version the adapter does not know yet, serve it through a compatible model the adapter does know. See Compatibility mode.
Further reading
- DIAL Vertex AI Adapter capabilities — thinking, safety settings, grounding, and caching
- DIAL Vertex AI Adapter configuration — Google Cloud credentials and every adapter environment variable
- Supported providers — models across all DIAL adapters