LLM Models
Supported AI model providers, model options, regional availability, BYOK, and model selection settings.
Overview
Workspace admins can configure AI model selection in Admin > Settings > AI Models where the workspace plan allows it. Model availability depends on the workspace plan, deployment region, provider availability, customer credentials, and customer agreement.
ContractRabbit separates model selection by use case:
| Use case | What it powers |
|---|---|
| Agent | Contract Q&A, tool use, drafting assistance, review workflows, and interactive analysis. |
| Knowledge extraction | Document metadata extraction, clause classification, field extraction, and structured contract data generation. |
| Embeddings | Semantic search, retrieval, clustering, matching, and related knowledge workflows where configured. |
Region support summary
| Region | Standard posture | Typical providers |
|---|---|---|
| United States | Standard managed production configuration. | Google Gemini / Vertex AI, OpenAI, Voyage for embeddings where configured. |
| European Union / EEA | Enterprise-scoped regional configuration where available. | Google Gemini / Vertex AI regional configuration, OpenAI EU data residency configuration where supported, Voyage only if provider terms and endpoint configuration satisfy the customer requirement. |
| China | Approved enterprise configuration only. | DeepSeek and Qwen / DashScope. |
EU/EEA, China, and other regional requirements should be confirmed during procurement because storage, processing, backups, support access, provider routing, and fallback behavior may all be in scope.
United States models
| Provider | Typical model | Use case | Notes |
|---|---|---|---|
| Google Gemini / Vertex AI | Gemini 3.1 Flash Lite | Agent | Default managed agent model. |
| Google Gemini / Vertex AI | Gemini 2.5 Flash Lite | Knowledge extraction | Default managed extraction model. |
| Google Gemini / Vertex AI | Gemini 3 Flash | Agent, knowledge extraction | Higher-capability Gemini option where enabled. |
| Google Gemini / Vertex AI | Gemini 2.5 Flash | Agent, knowledge extraction | General-purpose Gemini option. |
| Google Gemini / Vertex AI | Gemini 2.0 Flash | Agent, knowledge extraction | General-purpose Gemini option. |
| OpenAI | GPT-5 Mini | Agent, knowledge extraction | Typical OpenAI agent option and summary/document-generation option. |
| OpenAI | GPT-5 Nano | Knowledge extraction | Lower-cost OpenAI extraction option. |
| OpenAI | GPT-5 | Agent, knowledge extraction | Higher-capability OpenAI option where enabled. |
| Voyage | Embedding models | Embeddings | Used for retrieval and semantic search workflows where configured. |
European Union / EEA models
EU/EEA deployments require approved regional configuration. The model families below are typical enterprise options when the provider, model, endpoint, and retention terms support the customer's requirements.
| Provider | Typical model | Use case | EU deployment notes |
|---|---|---|---|
| Google Gemini / Vertex AI | Gemini 3.1 Flash Lite | Agent | Requires approved regional Vertex/Gemini configuration and model availability. |
| Google Gemini / Vertex AI | Gemini 2.5 Flash Lite | Knowledge extraction | Requires approved regional Vertex/Gemini configuration and model availability. |
| Google Gemini / Vertex AI | Gemini 3 Flash | Agent, knowledge extraction | Available only where the selected regional endpoint supports the model. |
| Google Gemini / Vertex AI | Gemini 2.5 Flash | Agent, knowledge extraction | Available only where the selected regional endpoint supports the model. |
| OpenAI | GPT-5 Mini | Agent, knowledge extraction | Requires approved OpenAI EU data residency configuration where supported. |
| OpenAI | GPT-5 Nano | Knowledge extraction | Requires approved OpenAI EU data residency configuration where supported. |
| OpenAI | GPT-5 | Agent, knowledge extraction | Requires approved OpenAI EU data residency configuration where supported. |
| Voyage | Embedding models | Embeddings | Available only if provider terms and endpoint configuration satisfy the residency requirement. |
For EU/EEA-restricted workspaces, model routing and fallback behavior are governed by the approved provider and region set in the applicable enterprise agreement or deployment plan.
China models
China-region model options are available only for approved enterprise deployments.
| Provider | Typical model | Use case | Notes |
|---|---|---|---|
| DeepSeek | DeepSeek V4 Pro | Agent | China-region agent option for approved deployments. |
| DeepSeek | DeepSeek V4 Flash | Agent, knowledge extraction | China-region extraction default for approved deployments. |
| Qwen / DashScope | Qwen 3.6 Plus | Agent | China-region agent alternate for approved deployments. |
| Qwen / DashScope | Qwen 3.5 Plus | Agent, knowledge extraction | China-region extraction alternate for approved deployments. |
China-region model use requires approved provider credentials, endpoint configuration, and enterprise review. Full-platform China residency requires a separate deployment review for hosting, storage, network, AI providers, transfers, backups, and support access.
BYOK and provider credentials
Bring your own key lets paid workspaces use customer-managed provider credentials where supported.
| Provider | BYOK support | Notes |
|---|---|---|
| Supported where plan and deployment allow it. | Customer is responsible for provider account configuration and applicable provider terms. | |
| OpenAI | Supported where plan and deployment allow it. | Customer is responsible for provider account configuration, data controls, and applicable provider terms. |
| DeepSeek | Enterprise-scoped. | Available only for approved China-region deployments. |
| Qwen / DashScope | Enterprise-scoped. | Available only for approved China-region deployments. |
Model selection controls
| Control | Description |
|---|---|
| Provider selection | Choose the approved provider for agent and extraction workloads. |
| Use-case model selection | Select separate models for agent workflows and knowledge extraction workflows. |
| AI Unit impact | Compare model cost impact through normalized AI Units instead of raw provider token pricing. |
| Credential mode | Use ContractRabbit-managed credentials or customer-managed credentials where the plan allows it. |
| Regional restrictions | Enterprise workspaces can define approved provider and region sets in the applicable agreement and deployment plan. |