> ## Documentation Index
> Fetch the complete documentation index at: https://phaseo.app/docs/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Phaseo Gateway base URL: https://api.phaseo.app/v1. Send `Authorization: Bearer $PHASEO_API_KEY` with every request.
> Model IDs look like `provider/model`, for example `openai/gpt-6-astra`. Do not guess IDs: list them with `GET /v1/models` or check https://phaseo.app/models. `phaseo/free` routes to an available free model and needs no wallet balance.
> For new text integrations prefer `POST /v1/responses`. `POST /v1/chat/completions` (OpenAI format) and `POST /v1/messages` (Anthropic format) suit existing code.
> The full API contract is https://phaseo.app/docs/openapi/v1/openapi.yaml. Official SDKs: `@phaseo/sdk` (TypeScript) and `phaseo` (Python), plus Go, C#, PHP, Ruby and Rust.
> Search these docs over MCP at https://phaseo.app/docs/mcp, and install Phaseo skills with `npx skills add https://phaseo.app/docs`. If a page is wrong or out of date, report it with the MCP `submit_feedback` tool.
> Manage Phaseo accounts with the `phaseo` CLI (`npm install -g @phaseo/cli`, then `phaseo login` or `phaseo login --device-code`), the `/v1` Management API or the MCP server at `https://mcp.phaseo.app/mcp?access=manage`. Do not send users to the dashboard for account tasks. A person is still needed for sign-up, approving a sign-in, paying the first top-up on the checkout link that `phaseo credits add` returns when no card is saved, confirming each saved-card top-up from the emailed link, and account security (password, MFA, passkeys). Create API keys with `phaseo keys create --name <name> --json`; never guess one. Credit top-ups need `phaseo login --scopes billing:write` first. Task map: https://phaseo.app/docs/v1/developers/manage-from-an-agent.md.

# OpenAI GPT-5.4

> What to know when adopting GPT-5.4, including reasoning controls, vision, and computer use.

# OpenAI GPT-5.4

Use this guide to adopt `openai/gpt-5.4` safely in production.

GPT-5.4 is less about a brand-new request shape and more about updated reasoning controls, vision handling, and tool use. As of March 10, 2026, OpenAI's official GPT-5.4 docs emphasize `reasoning.effort`, improved computer use, and higher-fidelity image input. They do not document a GPT-5.4-specific `phase` parameter.

## Migration quickstart

1. Change the model ID to `openai/gpt-5.4` without changing the rest of the request.
2. Choose and test the reasoning effort that matches the route's latency budget.
3. Re-test vision, structured outputs, and computer-use workflows where applicable.
4. Canary GPT-5.4 with the previous route available as a fallback.

## What's new

* `reasoning.effort` supports `none`, `low`, `medium`, `high`, and `xhigh`
* GPT-5.4 introduces `original` image input detail for full-fidelity visual perception
* the updated `computer` tool is the official path for GPT-5.4 computer-use workflows
* GPT-5.4 improves long-horizon tool use, large-context work, and professional document workflows

## What to change in your integration

### 1. Review reasoning settings

If you are coming from GPT-4.1 or an earlier GPT-5 variant, do not assume your previous reasoning defaults are still the best choice. Re-test routes at the effort level that matches their latency budget:

* `none` or `low` for low-latency routes
* `medium` or `high` for most structured tasks
* `xhigh` for deeply reasoned or agentic workflows

### 2. Revisit vision pipelines

If your workflow depends on dense screenshots, documents, or UI localization, test the new `original` image detail level. This changes both fidelity and cost characteristics.

### 3. Re-check computer-use routes

If you are migrating browser or desktop automation tasks, update them to the current `computer` tool path and retest step completion end to end.

## What to test

### Prompt compatibility

* long system prompts
* extraction and classification prompts
* format-sensitive outputs
* prompts that previously needed retries to stay on task

### Tools and structured outputs

* schema adherence at each effort level
* tool selection quality
* argument completeness
* stability of computer-use flows across multiple steps

### Vision and operations

* screenshot and document understanding with `high` versus `original` detail
* latency and token usage under realistic load
* retry and timeout rate
* cost per successful task versus the model you are replacing

## Safe rollout

1. Benchmark GPT-5.4 with the exact effort levels you intend to use.
2. If you use images, test `original` detail separately because it changes both quality and cost.
3. Shadow production prompts before switching any default route.
4. Promote gradually with an older OpenAI model or alternate provider still available as fallback.

## Sources

* [Introducing GPT-5.4](https://openai.com/index/introducing-gpt-5-4/)
* [GPT-5.4 model docs](https://developers.openai.com/api/docs/models/gpt-5.4)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.