Supported OpenAI models for AI integration
Introduced in 2024.2
Updated in 2025
This article explains which AI models you need to configure for the AI integration in Alloy Navigator and what each model is used for. You select models during AI integration, when entering your API key and configuring AI features.
Alloy Navigator supports OpenAI API and Azure OpenAI Service AI providers. Both offer access to powerful AI models, with OpenAI API offering a quick and flexible setup, while Azure adds enterprise-grade features like regional deployment, compliance options, and deep integration with Microsoft services.
IMPORTANT: If you use Azure OpenAI Service, you first need to create a deployment for each model you want to use. For details, see Azure documentation. Then, you specify the deployment name in Alloy Navigator.
Use this table for a quick overview and to explore details as needed.
| Model | Required or optional | Description | AI feature usage |
|---|---|---|---|
|
Required |
Processes a prompt and produces a context-aware response |
||
|
Optional |
Converts written text into spoken words |
||
|
Optional |
Transcripts spoken language into text |
||
|
Optional |
Generates vector representations of text for semantic search and internal knowledge access |
INFO: For a full listing of available OpenAI models, see https://platform.openai.com/docs/models. For information about Azure OpenAI Service models, see https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models.
Completion model (Chat, Language, Text Generation model)
The completion model (also called a chat, language or a text generation model), which completes the initial input ("prompt") by providing a plausible continuation, is essential for all AI features in Alloy Navigator, including AI Writing Assistance and AI-Powered Insights. Completion models are used for a wide variety of tasks, including text generation, question answering, or translating text from one language to another.
GPT models are fast, cost-efficient, and highly intelligent, and work well with the clear, detailed instructions already built into our workflows. By default, we use GPT-4o mini, a chat model optimized for conversational AI, but you can consider alternatives, as OpenAI suggests. Mini and nano models offer trade-offs in speed and cost, generally being faster and cheaper to use. Since new models are regularly released, we support both selecting from the dropdown menu and entering a model name manually to keep you up to date.
If you use Azure OpenAI Service, choose a GPT model that can understand and generate natural language.
IMPORTANT: Avoid selecting reasoning models. Unlike chat models that generate direct answers, reasoning models, such as o3, provide step-by-step thinking and explanation, which don't fit into Alloy Navigator AI-assisted workflows.
Text-to-Speech model
AI Writing Assistance uses the Text-to-Speech model to convert written text into speech imitating human voice. By default, we use tts-1, but you can select or enter a different model. If you use Azure OpenAI Service, the model type is Audio.
Using this model is optional. If you don't plan to use the Text-to-Speech feature, you can disable it. AI Writing Assistance will continue to work, just without voice output.
Transcription model
The Transcription model is used for speech recognition in AI Writing Assistance, converting audio to written text. The default model is whisper-1. If you use Azure OpenAO Service, the model type is Audio.
Using the transcription model is optional. If you don't plan to use the speech recognition feature, you can disable it. AI Writing Assistance will continue to work, just without voice-to-text transcription.
Embedding model
The embedding model generates vector representations of text for semantic search and internal knowledge access for AI-Powered Insights. The default model is text-embedding-3-large. If you use Azure OpenAO Service, the model type is Embeddings.
Using the embedding model is optional. If you decide not to use it and disable access to internal knowledge, AI-Assisted Insights will still work but provide less accurate and relevant responses because it won't be able to search your past tickets or knowledge base.