GPT-5.2, refreshed in December 2025, offers Atlas users a robust general-purpose model with a 400K token context window and a 128K max output ceiling. It is well-suited for read-many, write-large agent turns within Atlas, despite its $1.75 per Mtok input and $14 per Mtok output pricing being higher than its predecessor.
What is GPT-5.2 best at for Atlas?
GPT-5.2, a December 2025 refresh, excels within Atlas for tasks requiring extensive context and substantial output, leveraging its 400K token context window. This model is a strong general-purpose choice for developers in 2026, especially for complex coding agent turns.
GPT-5.2 from OpenAI is a frontier model, ideal for driving Atlas, the terminal-native AI coding agent. Its 400K token context window, paired with a 128K max output ceiling, provides a workable ratio for agent turns that involve reading large codebases and generating significant output. As the general purpose sibling of GPT-5.2 Codex, it offers flexibility for a wide range of coding tasks, allowing developers to swap between them on the same API key and context window. Atlas connects to Model Context Protocol servers and exposes their tools to the agent, making GPT-5.2's reasoning capabilities accessible through the Responses API. This enables Atlas to draft plans in a read-only plan agent, ask for approval, and then switch to a build agent, computing unified diffs for every file edit before writing. Atlas also uses Axis, the hybrid semantic and keyword code search, to efficiently retrieve relevant code, which benefits from GPT-5.2's large context.
What are the cost and context tradeoffs of GPT-5.2?
GPT-5.2 offers a substantial 400K token context window, but developers in 2026 must consider its increased pricing: $1.75 per Mtok input and $14 per Mtok output. This represents a 40% increase in input cost compared to GPT-5.1.
While GPT-5.2 maintains the impressive 400K token context window and 128K max output ceiling of its predecessor, GPT-5.1, its pricing has increased. Input tokens are billed at $1.75 per Mtok, and output tokens at $14 per Mtok. This makes GPT-5.2 40% more expensive per input token than GPT-5.1, which was priced at $1.25 per Mtok input. Developers using Atlas need to weigh this increased cost against the model's general-purpose reasoning capabilities and large context window. For scenarios where cost optimization is paramount, especially for frequent or high-volume tasks, this pricing structure requires careful consideration. Atlas allows switching the active model and provider on the fly, enabling developers to compare GPT-5.2's performance and cost against other models for specific tasks.
When should I pick a different model over GPT-5.2 for Atlas?
For pure agentic coding tasks within Atlas, GPT-5.2 Codex is often a better choice than GPT-5.2, despite both models sharing the same 400K context window and pricing. GPT-5.2 is not post-trained specifically for coding, unlike its specialized sibling.
While GPT-5.2 is a capable general-purpose model for Atlas, developers should consider GPT-5.2 Codex for tasks that are exclusively focused on code generation and manipulation. GPT-5.2 Codex is post-trained for coding, making it usually the better pick for pure agentic coding scenarios at an identical price point and with the same 400K token context window and 128K max output. Atlas allows developers to use /models to compare models head-to-head on the same repository task, facilitating an informed decision. Additionally, for very cheap calls or specific subagent tasks, setting "small_model": "openai/gpt-5-nano" can prevent billing at the full $14 per Mtok output rate of GPT-5.2, offering a cost-effective alternative for less demanding operations. Atlas's ability to fan out work to subagents, which can run in the foreground or in parallel background sessions, further supports using specialized or cheaper models for specific parts of a larger task.
Setup
- 01Set OPENAI_API_KEY in the environment.
- 02Run `atlas models openai` to confirm gpt-5.2 resolves from models.dev.
- 03Pin "model": "openai/gpt-5.2" in atlas.json.
- 04Set "small_model": "openai/gpt-5-nano" so cheap calls do not bill at $14 per Mtok output.
- 05Use /models to compare it head to head with gpt-5.2-codex on the same repo task.
Frequently asked questions
- What is the context window size for GPT-5.2 in Atlas?
- GPT-5.2 provides a substantial 400K token context window for use within Atlas, enabling the agent to process extensive codebases and documentation.
- How much does it cost to use GPT-5.2 with Atlas?
- Using GPT-5.2 with Atlas costs $1.75 per Mtok for input tokens and $14 per Mtok for output tokens. This pricing is a key consideration for developers in 2026.
- Is GPT-5.2 better than GPT-5.2 Codex for coding in Atlas?
- For pure agentic coding tasks, GPT-5.2 Codex is typically the better choice as it is post-trained for coding, while GPT-5.2 is a general-purpose model. Both share the same 400K context window and pricing.
- Can I switch between GPT-5.2 and other models in Atlas?
- Yes, Atlas allows you to switch the active model and provider on the fly, with support for favorites and recents, making it easy to compare GPT-5.2 with other options.
- What is the maximum output for GPT-5.2 in Atlas?
- GPT-5.2 has a 128K token maximum output ceiling, which is a workable ratio for read-many, write-large agent turns within Atlas.
- How does GPT-5.2's pricing compare to GPT-5.1?
- GPT-5.2 is 40% more expensive per input token than GPT-5.1, with input tokens priced at $1.75 per Mtok compared to GPT-5.1's $1.25 per Mtok.
- How can I optimize costs when using GPT-5.2 with Atlas?
- To optimize costs, consider setting "small_model": "openai/gpt-5-nano" in your atlas.json for cheap calls, preventing them from billing at the full $14 per Mtok output rate of GPT-5.2.
Try SeaShell in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install SeaShellRelated guides
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.
Atlas vs Kiro: Terminal AI Coding Agents in 2026
Compare Atlas, the terminal-native AI coding agent, with Kiro, AWS's spec-driven IDE and CLI, for developers in 2026. Evaluate pricing, workflows, model flexibility, and code safety.
Atlas for Laravel in 2026
Atlas is a terminal-native AI coding agent for Laravel in 2026. Run it in a Laravel app with an artisan file, add a controller or Pest test, and review the diff.
Plan a Multi-File Change Before Editing with Atlas in 2026
How to plan a multi-file change with Atlas in 2026: the plan agent denies all edit tools, you research with codebase_search and lsp, then plan_exit hands off.
Add a Regression Test for a Bug Fix with Atlas in 2026
How to add a regression test with Atlas in 2026: red first, then green. bash records the exit code, write creates the failing test, and edit applies the fix.
Atlas for Dart in 2026
Adopt Atlas, the terminal-native AI coding agent, for Dart development in 2026. Enhance productivity with intelligent code search, refactoring, and robust safety features across your Dart projects.
Atlas vs Fine.dev: Terminal AI Coding Agents in 2026
Atlas and Fine.dev offer distinct AI coding agent workflows for 2026. Atlas provides a terminal-native TUI with permission-gated tool calls, while Fine.dev focuses on asynchronous cloud-based agent execution and PR
Atlas for Expo: Terminal-Native AI Coding for expo-router and Config Plugins in 2026
Atlas is a terminal-native AI coding agent for Expo apps in 2026, covering expo-router file routes, config plugins, and EAS build profiles with diff-first review.