Azure OpenAI (gateway) puts the whole GPT-5 family, including the Codex line, under Azure IAM and your existing enterprise agreement, and Atlas resolves Azure deployments through the Responses API, falling back to chat completions. Context varies by model, up to 1.05M on the GPT-5.4+ line, and pricing is passthrough at Azure rates, broadly matching OpenAI's direct pricing. The friction is deployments: you must create a named deployment per model in the Azure portal before Atlas can call it.
What is Azure OpenAI best at inside Atlas?
Azure OpenAI (gateway) is best inside Atlas at delivering the coding-specialized GPT-5 models to an enterprise. The full GPT-5 lineup on Azure includes gpt-5.3-codex, gpt-5.1-codex-max, and gpt-5.1-codex-mini, so the Codex models are not enterprise-gated behind a separate agreement.
The Codex line is the reason to reach for Azure OpenAI (gateway) inside Atlas rather than any other enterprise gateway. gpt-5.3-codex and gpt-5.1-codex-max are tuned for exactly what Atlas asks a model to do: read AST-indexed declarations returned by hybrid semantic and keyword retrieval, plan a change, and emit edits that Atlas turns into a unified diff for approval before writing. gpt-5.1-codex-mini is the cheap slot for titles and summaries. All three run inside your Azure region under your existing enterprise agreement, so the security review that already covered Azure covers them.
Why does Atlas need an Azure deployment before it sees a model?
Azure OpenAI (gateway) requires a named deployment per model in the Azure portal before Atlas can call it, so a GPT-5 model existing in the catalog is not enough. Create the deployment, then run `atlas models azure` to confirm Atlas actually sees it.
Azure's deployment model is the single most common Azure OpenAI (gateway) surprise for Atlas users. On a direct API, a model ID is enough. On Azure, a model is a catalog entry and a deployment is an instance of it in your resource, and only the deployment is callable. If `atlas models azure` shows nothing, the fix is in the Azure portal, not in atlas.json. Once the deployment exists, Atlas resolves it, and Atlas's Azure loader tries `sdk.responses(modelID)` first and falls back to `sdk.chat(modelID)`, so older deployments keep working alongside new ones.
How do you point Atlas at an Azure OpenAI resource?
Point Atlas at Azure OpenAI (gateway) with 2 environment variables: set `AZURE_RESOURCE_NAME=my-resource` and `AZURE_API_KEY=...`, and Atlas builds the baseURL as `https://<resource>.cognitiveservices.azure.com/openai`. If your resource sits behind a different endpoint, set an explicit `baseURL` in the azure provider block in atlas.json.
Two environment variables get Azure OpenAI (gateway) working with Atlas in most cases: `AZURE_RESOURCE_NAME` and `AZURE_API_KEY`. Atlas derives the endpoint from the resource name, producing `https://<resource>.cognitiveservices.azure.com/openai`. Private endpoints, custom domains, and non-default sovereign clouds all break that derivation, which is why the azure provider block in atlas.json accepts an explicit `baseURL` override. Set it once and the rest of the flow is identical: create the deployment, run `atlas models azure`, and select the model from `/models`. Atlas lets you switch the active model and provider on the fly with favorites and recents.
What context window does Azure OpenAI give Atlas?
Context on Azure OpenAI (gateway) varies by model, up to 1.05M on the GPT-5.4+ line. A 1.05M window is enough for Atlas to keep a wide hybrid search result, several files, and an accumulating stack of unified diffs resident through a long refactor without dropping context.
Because Azure OpenAI (gateway) is passthrough at Azure rates, broadly matching OpenAI's direct pricing, the large windows on the GPT-5.4+ line cost the same as they do direct. Inside Atlas, that window is what lets a long autonomous session stay coherent. Atlas fans out work to subagents that can run in the foreground or in parallel background sessions, and Atlas snapshots file changes as git patches so edits can be diffed and rolled back. With 1.05M tokens, the parent session can hold the full picture while subagents work. Confirm the window your specific deployment exposes with `atlas models azure`.
When should you pick the direct OpenAI API instead of Azure?
Pick the direct OpenAI API over Azure OpenAI (gateway) when you need the newest model immediately. New OpenAI models reach Azure weeks after the direct API, so the newest GPT-5.6 tier may not be available on Azure yet, and every model needs a named deployment first.
The honest calculus on Azure OpenAI (gateway). The reason to choose it is that data stays in your chosen Azure region under your existing enterprise agreement, which is usually the entire reason for choosing it. If that does not apply to you, Azure is pure overhead: a deployment per model, a resource name, and a lag of weeks on new releases. If it does apply, the overhead is the price of shipping at all, and the GPT-5 Codex line is fully available. Atlas supports both, so a team can run Azure for work repositories and the direct API for open-source side projects.
Setup
- 01Set `AZURE_RESOURCE_NAME=my-resource` and `AZURE_API_KEY=...`. Atlas builds the baseURL as `https://<resource>.cognitiveservices.azure.com/openai`.
- 02Alternatively, set an explicit `baseURL` in the azure provider block in atlas.json to override the derived one.
- 03Create a deployment for the model in the Azure portal, then run `atlas models azure` to confirm Atlas sees it.
- 04Select it from `/models` in the Atlas TUI.
Frequently asked questions
- how to connect atlas to azure openai
- Set `AZURE_RESOURCE_NAME=my-resource` and `AZURE_API_KEY=...`. Atlas builds the baseURL as `https://<resource>.cognitiveservices.azure.com/openai`. Create the model deployment in the Azure portal, run `atlas models azure`, then select it from `/models`.
- why does atlas models azure show no models
- You must create a named deployment per model in the Azure portal before Atlas can call it; the model existing in the catalog is not enough. Create the deployment, then re-run `atlas models azure`.
- can i use gpt-5 codex models on azure openai
- Yes. The full GPT-5 lineup on Azure includes gpt-5.3-codex, gpt-5.1-codex-max, and gpt-5.1-codex-mini, so the coding-specialized models are not enterprise-gated behind a separate agreement.
- how much does azure openai cost compared to openai direct
- Azure OpenAI is passthrough at Azure rates, broadly matching OpenAI's direct pricing. The value of Azure is that data stays in your chosen Azure region under your existing enterprise agreement, not a lower price.
- what is the context window on azure openai
- Context on Azure OpenAI varies by model, up to 1.05M on the GPT-5.4+ line. Run `atlas models azure` to confirm what your specific deployment exposes before pinning it.
- how do i override the azure baseurl in atlas
- Set an explicit `baseURL` in the azure provider block in atlas.json to override the one Atlas derives from `AZURE_RESOURCE_NAME`. Use this for private endpoints, custom domains, or sovereign clouds.
- is the newest openai model available on azure
- Not always. New OpenAI models reach Azure weeks after the direct API, so the newest GPT-5.6 tier may not be available yet. Use the direct OpenAI API if day-one access to a new release matters.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas vs PearAI: Choosing Your AI Coding Agent in 2026
Compare Atlas, the terminal-native AI coding agent, with PearAI, a VS Code fork. Evaluate features, pricing, and community support for developers in 2026.
Rename a Symbol Across the Repo with Atlas in 2026
How to rename a symbol across a repo with Atlas in 2026: findReferences gets the true reference set, grep catches strings and docs, and edit refuses ambiguous matches.
Atlas vs Cosine: Terminal AI Coding Agents in 2026
Comparing Atlas and Cosine for terminal AI coding in 2026. Atlas offers a free core and local privacy, while Cosine provides proprietary models and a cloud surface.
Atlas for Astro: Islands, Content Collections, and Zero JS by Default in 2026
Atlas is a terminal-native AI coding agent for Astro in 2026. It reads astro.config.mjs, src/pages, and content collection schemas, drops needless client:load directives, and runs astro check.
Atlas vs Sourcery: Terminal AI Coding Agents in 2026
Compare Atlas and Sourcery for 2026. Atlas offers terminal-native AI coding with permission-gated tools and local embeddings. Sourcery provides Python-first LLM review and rules-based refactoring, starting at
Atlas vs Graphite: Terminal AI Coding Agents in 2026
Comparing Atlas and Graphite in 2026: Atlas offers a terminal-native AI coding agent with local privacy, while Graphite focuses on stacked PR workflows and GitHub integration.
Atlas for C++ in 2026
In 2026, C++ developers adopt Atlas, the terminal-native AI coding agent, to enhance productivity. Atlas offers secure, context-aware assistance for modern C++ projects, integrating with CMake and ensuring code quality
Atlas for dbt: Terminal-Native AI Coding in 2026
Atlas is a terminal-native AI coding agent for dbt. Read the ref() DAG, convert a table model to incremental, run dbt build against dev, and add tests in 2026.