[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-09-04。"],[],[],null,["# Overview of self-deployed models\n\nModel Garden offers both self-deployed open, partner, and custom models\nthat you can deploy and serve on Vertex AI. These models are different\nfrom the [model-as-a-service (MaaS)](/vertex-ai/generative-ai/docs/partner-models/use-partner-models) offerings, which are serverless and\nrequire no manual deployment.\n\nWhen you self deploy models, you deploy them securely within your\nGoogle Cloud project and VPC network.\n\nSelf-deploy open models\n-----------------------\n\nOpen models provide pretrained capabilities for various AI tasks, including\nGemini models that excel in multimodal processing. An open model is\nfreely available, you are free to publish its outputs, and it can be used\nanywhere as long as you adhere to its licensing terms.\n[Vertex AI](/vertex-ai) offers both open (also known as *open weight*)\nand open source models.\n\nWhen you use an open model with Vertex AI, you use Vertex AI for\nyour infrastructure. You can also use open models with other infrastructure\nproducts, such as PyTorch or Jax.\n\n### Open weight models\n\nMany open models are considered open weight large language models (LLMs). Open\nmodels provide more transparency than models that aren't open weight. A\nmodel's weights are the numerical values stored in the model's neural network\narchitecture that represent learned patterns and relationships from the data a\nmodel is trained on. The pretrained parameters, or weights, of open weight\nmodels are released. You can use an open weight model for inference and tuning\nwhile details such as the original dataset, model architecture, and training\ncode aren't provided.\n\n### Open source models\n\nOpen models differ from open source AI models. While open models often expose\nthe weights and the core numerical representation of learned patterns, they\ndon't necessarily provide the full source code or training details. Providing\nweights offers a level of AI model transparency, allowing you to\nunderstand the model's capabilities without needing to build it yourself.\n\nSelf-deployed partner models\n----------------------------\n\n|\n| **Preview**\n|\n|\n| This feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n\nModel Garden helps you purchase and manage model licenses from partners\nwho offer proprietary models as a self deploy option. After you purchase\naccess to a model from Cloud Marketplace, you can choose to deploy on\non-demand hardware or use your Compute Engine reservations and\ncommitted use discounts to meet your budget requirements. You are charged for\nmodel usage and for the Vertex AI infrastructure that you use.\n\nTo request usage of a self-deployed partner model, find the relevant model in\nthe [Model Garden console](https://console.cloud.google.com/vertex-ai/model-garden), click **Contact sales**, and\nthen complete the form, which initiates contact with a Google Cloud sales\nrepresentative.\n\nFor more information about deploying and using partner models, see [Deploy a\npartner model and make prediction requests](/vertex-ai/generative-ai/docs/model-garden/use-models#deploy_a_partner_model_and_make_prediction_requests).\n\n### Considerations\n\nConsider the following limitations when using self-deployed partner models:\n\n- Unlike with open models, you cannot export weights.\n- If you VPC Service Controls set up for your project, you can't upload models, which prevents you from deploying partner models.\n- For endpoints, only the [shared public endpoint](/vertex-ai/docs/predictions/choose-endpoint-type) type is supported.\n\n| **Note:** Support for model-specific issues is provided by the partner. To contact a partner for model performance related issues, use the contact details in the \"Support\" section of their Model Garden model card.\n\nLearn more about self-deployed models in Vertex AI\n--------------------------------------------------\n\n- To learn more about custom weights, see [Deploy models with custom\n weights](/vertex-ai/generative-ai/docs/model-garden/deploy-models-with-custom-weights).\n- For more information about Model Garden, see [Overview of\n Model Garden](/vertex-ai/generative-ai/docs/model-garden/explore-models).\n- For more information about deploying models, see [Use models in\n Model Garden](/vertex-ai/generative-ai/docs/model-garden/use-models).\n- [Use Gemma open models](/vertex-ai/generative-ai/docs/open-models/use-gemma)\n- [Use Llama open models](/vertex-ai/generative-ai/docs/open-models/use-llama)\n- [Use Hugging Face open models](/vertex-ai/generative-ai/docs/open-models/use-hugging-face-models)"]]