OCI Generative AI

What it does

OCI Generative AI is a managed Oracle Cloud Infrastructure service for building and running generative AI applications. It takes in prompts, documents, queries, and custom model artifacts, and produces chat responses, embeddings, relevance-ranked results, and agent outputs. Access runs through the Console, API, CLI, and OpenAI-compatible endpoints.

Teams use it to call pretrained hosted models for chat, semantic search, and reranking, or to import, fine-tune, and host custom models on dedicated AI clusters. For agents, the OCI Responses API provides tool use such as File Search, Code Interpreter, Function Calling, and MCP Calling, alongside memory and NL2SQL for structured enterprise data. Hosted agentic applications can run as containers on OCI-managed infrastructure.

It suits developers and enterprise teams already on Oracle Cloud who need IAM policies, private endpoints, and guardrails around AI use. Using it requires an Oracle Cloud Infrastructure account, and dedicated clusters carry minimum unit-hour commitments. Pricing is metered per token, per unit-hour, or per storage and request, so costs depend on usage and should be checked against Oracle's rate table.

What it can do

• Chat generation with pretrained hosted models for conversational experiences
• Embeddings for semantic search, recommendation, classification, and clustering
• Rerank to order documents by relevance to a query
• OpenAI-compatible APIs for use with existing tools and SDKs
• Dedicated AI clusters to host or fine-tune custom models
• Agent building with the OCI Responses API, including File Search, Code Interpreter, Function Calling, and MCP Calling

Works with

• OpenAI-compatible API
• MCP Calling

Example workflows

Semantic document search

Illustrative example. Input: a document collection and user queries. Steps: create embeddings for the documents, store them in a vector store, then use rerank to order retrieved results by relevance to each query. Output: a relevance-ordered list of matching documents for the search interface.

Enterprise data question answering agent

Illustrative example. Input: a natural-language question about structured enterprise data. Steps: an agent built on the OCI Responses API uses SQL Search (NL2SQL) to convert the request into validated SQL, runs it, and applies guardrails to the response. Output: an answer grounded in the queried data.

Before you choose

• Dedicated deployments have model- and resource-specific commitments; check the current service limits and rate table before provisioning.
• Pricing is metered by tokens, AI unit-hours, storage, and requests, so the total depends on usage and the price-list rate table.
• Governance features such as private endpoints, guardrails, and ZPR are tied to OCI networking and identity setup, which adds configuration effort.

Listed under

• Coding

Get started

An OCI account, supported region and IAM policies are required.

1. Create or sign in to an Oracle Cloud Infrastructure account and set up IAM policies for who can use Generative AI resources.
2. Choose a pretrained model for chat or embeddings, then call it through the Console, API, or CLI, or generate an API key.
3. For custom models or steady traffic, create a dedicated AI cluster with the required AI units and minimum unit-hour commitment.

Hosted model calls and dedicated resources have separate meters. Consult the relevant Generative AI rows in Oracle’s cloud price list; storage, requests and agent services can add charges. General cloud trial credits are not a permanent free model allowance.

Source: oracle.com/cloud/price-list, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• OCI Generative AI Overview ↗docs.oracle.com
• Cloud Price List | Oracle ↗www.oracle.com

Company website

docs.oracle.com: product, pricing and policy pages

Web archive

When the site first appeared

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.