GPT-5.4 brings native computer-use capabilities, tool search for token efficiency, and GPT-5.4 Thinking in ChatGPT, API, and Codex.
No credit card required
Source-backed capabilities from OpenAI’s official release
OpenAI launched GPT-5.4 in ChatGPT as “GPT-5.4 Thinking,” with API access via gpt-5.4 and integration in Codex workflows.
OpenAI describes GPT-5.4 as its first general-purpose model with native computer-use capabilities for multi-app workflows.
Instead of preloading all tools into context, GPT-5.4 can look up tool definitions on demand. OpenAI reports up to 47% token reduction in testing.
Thinking mode can present an upfront plan and supports user guidance while responses are in progress for more controllable execution.
GPT-5.4 supports up to 1M tokens in Codex experimental settings for large-scale repositories and long technical contexts.
OpenAI reports 83.0% on GDPval and 75.0% on OSWorld-Verified, positioning GPT-5.4 for high-stakes professional tasks.
As published by OpenAI
| Benchmark | GPT-5.4 | Context |
|---|---|---|
| GDPval (professional work comparisons) | 83.0% | Matching or exceeding industry professionals |
| OSWorld-Verified (computer use) | 75.0% | Surpasses reported human benchmark (72.4%) |
| SWE-Bench Pro | 57.7% | Reported on the launch page |
Source: OpenAI — Introducing GPT-5.4
From the official GPT-5.4 announcement page
| Model | Input | Output |
|---|---|---|
| gpt-5.4 | $2.50 / 1M | $15 / 1M |
| gpt-5.4-pro | $30 / 1M | See OpenAI pricing page |
Get started in minutes
Install globally with npm. Works on macOS, Linux, and Windows.
Create your account — takes less than a minute.
Use /model to pick GPT-5.4 Thinking from the model selector.
Run GPT-5.4 Thinking in AdaL CLI with model switching, tool orchestration, and production-ready coding workflows.
Get Started with AdaL