Oracle Cloud Infrastructure Documentation

Compatible Alibaba Models

You can import large language models from Hugging Face and OCI Object Storage buckets into OCI Generative AI, create endpoints for those models, and use them in the Generative AI service.

Alibaba Qwen model family, feature advanced multilingual and multimodal capabilities. For model cards on Hugging Face, see the links in the following tables.

Qwen 3 Next

Compatible Qwen 3 Next Model
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen3-Next-80B-A3B-Instruct TEXT_TO_TEXT
  • A100_80G_X4
  • H100_X4
  • H200_X4

Qwen 3.6

Compatible Qwen 3.6 Model
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen3.6-35B-A3B IMAGE_TEXT_TO_TEXT
  • A100_80GB_X2
  • H100_X2
  • H200_X2
  • B200_X2

Qwen 3.5

Compatible Qwen 3.5 Model
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen3.5-9B IMAGE_TEXT_TO_TEXT
  • A100_80GB_X1
  • H100_X1
  • H200_X1
  • B200_X1

Qwen Image

Compatible Qwen Image Models
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen-Image TEXT_TO_IMAGE A100_80G_X1
Qwen/Qwen-Image-Edit IMAGE_TEXT_TO_IMAGE A100_80G_X1
Qwen/Qwen-Image-2512 TEXT_TO_IMAGE A100_80G_X1
Qwen/Qwen-Image-Edit-2511 IMAGE_TEXT_TO_IMAGE A100_80G_X1
Qwen/Qwen-Image-Edit-2509 IMAGE_TEXT_TO_IMAGE A100_80G_X1
Note

  • response_format: "url" doesn't work and returns an HTTP 400 bad request error.
  • n (number of images): only 0 or 1 work.
  • Streaming isn’t compatible.
  • Non-standard image sizes might be rounded (for example, 999x999 → 992x992) instead of returning an HTTP 400 (unlike the OpenAI API).
  • Transparency might not work because of model limitations.

Qwen Q (Reasoning)

Compatible Qwen Q Models
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/QwQ-32B TEXT_TO_TEXT A100_80G_X2

Qwen 3

Compatible Qwen 3 Models
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen3-Embedding-0.6B EMBEDDING A10_X1
Qwen/Qwen3-Embedding-4B EMBEDDING A10_X2
Qwen/Qwen3-Embedding-8B EMBEDDING A100_80G_X1
Qwen/Qwen3-0.6B TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen3-1.7B TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen3-4B TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen3-8B TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen3-14B TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen3-32B TEXT_TO_TEXT A100_80G_X2
Qwen/Qwen3-4B-Instruct-2507 TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen3-30B-A3B-Instruct-2507 TEXT_TO_TEXT A100_80G_X2
Qwen/Qwen3-235B-A22B-Instruct-2507 TEXT_TO_TEXT H100_X8
Qwen/Qwen3-VL-30B-A3B-Instruct IMAGE_TEXT_TO_TEXT H100_X2
Qwen/Qwen3-VL-30B-A3B-Instruct-FP8 IMAGE_TEXT_TO_TEXT
  • H100_X2
  • H200_X2
Qwen/Qwen3-VL-235B-A22B-Instruct IMAGE_TEXT_TO_TEXT H100_X8

Qwen 2.5

Compatible Qwen2.5 Models
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen2.5-Coder-32B-Instruct TEXT_TO_TEXT A100_80G_X2
Qwen/Qwen2.5-0.5B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-1.5B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-3B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-7B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-14B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-32B-Instruct TEXT_TO_TEXT A100_80G_X2
Qwen/Qwen2.5-72B-Instruct TEXT_TO_TEXT A100_80G_X4
Qwen/Qwen2.5-VL-3B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-VL-7B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2.5-VL-32B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X2
Qwen/Qwen2.5-VL-72B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X4

Qwen 2

Compatible Qwen2 Models
Hugging Face Model ID Model Capability Recommended Dedicated AI Cluster Unit Shape
Qwen/Qwen2-0.5B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2-1.5B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2-7B-Instruct TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2-72B-Instruct TEXT_TO_TEXT A100_80G_X4
Qwen/Qwen2-VL-2B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2-VL-7B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X1
Qwen/Qwen2-VL-72B-Instruct IMAGE_TEXT_TO_TEXT A100_80G_X4
Important

  • For imported models, you can use the native context length specified by the model provider. However, the effective maximum context length is limited by the underlying hardware setup that you select for the hosting dedicated AI clusters in OCI Generative AI. To take full advantage of a model's native context length, you might need to provision more hardware resources.
  • Use the fine-tuned models only if they match the compatible base model's transformer version and have a parameter count within ±10% of the original.
  • For available hardware and steps on how to deploy the imported models, see Managing Imported Models.