OCI GenAI Model Catalog: A Reference Guide for Enterprise Generative AI Selection

This article introduces the open-source project oci-genai-catalog, a reference catalog that comprehensively organizes models from Oracle Cloud Infrastructure (OCI) Generative AI services. Hosted on GitHub Pages, the project aims to address the issue of scattered OCI official documentation by providing users with a unified, real-time updated platform for querying model information. The catalog covers native models from major vendors like Cohere, Google, Meta, OpenAI, and xAI, as well as 84 importable models, helping enterprises quickly compare specifications and make selection decisions.

With the development of generative AI technology, enterprises face challenges in model selection: dozens of models have their own advantages and disadvantages, and OCI official documentation is scattered across different pages, requiring developers to spend a lot of time organizing information. The oci-genai-catalog project was created to address this pain point, providing a single, comprehensive reference view of OCI Generative AI models.

Model Categories

The catalog covers four categories: conversational models, embedding models, reranking models, and importable models.

Native Provider Models

• Cohere: Command series (inference/vision, etc.), Embed v4/v3, Rerank 4.0, etc.;
• Google: Gemini 2.5 Pro/Flash/Flash-Lite (multimodal, million-level context);
• Meta: Llama4 (Maverick/Scout), Llama3.3/3.2/3.1 series (open-source and commercializable);
• OpenAI: gpt-oss-120b/20b (open-source weight models);
• xAI: Grok 4.3/4/Fast series (real-time information acquisition).

Importable Models

It收录 84 compatible models, including Alibaba Qwen series, DeepSeek, Google Gemma, Microsoft Phi, Mistral, NVIDIA Nemotron, OpenAI Whisper, etc.

Selection Guide

A four-step guide helps users quickly filter models: Use case identification → Trade-off between quality and speed → Deployment method selection → Region selection.

Detailed Specification Comparison

Each model entry includes key information such as Model ID, Tier, Context Window, Multimodal, Tool Use, Fine-tuning, Reasoning, Status, Best For.

Other Features

• Support dark/light theme switching;
• Pure static HTML/CSS/JS architecture: ultra-fast loading, simple deployment, low maintenance cost, mobile-friendly.

Data Source

All data comes from OCI official documentation, with a verification date of June 2026 to ensure accuracy. Currently, it mainly covers commercial OCI regions (OC1), and sovereign cloud and government regions are not included.

Applicable Users

Cloud architects, development engineers, product managers, enterprise decision-makers, AI researchers.

Typical Process

Explore → Filter → Compare → Verify → Decide.

Current Limitations

1. Data updates rely on manual work, with potential lag risks;
2. Lack of pricing information;
3. Limited regional coverage;
4. No performance benchmark data.

Improvement Directions

1. Automated data synchronization scripts;
2. Add price calculator;
3. Introduce community reviews;
4. Integrate performance benchmark tests;
5. Support multi-language interface.

oci-genai-catalog is an open-source project that precisely addresses the pain points of OCI users in model selection. Its success lies in focusing on the OCI platform, complete information, toolized design, and minimalist implementation. For enterprises and developers using or planning to use OCI Generative AI services, this project is an invaluable reference tool and also contributes to the improvement of the cloud ecosystem.