GPT-5.4
GPT-5.4 by OpenAI is the first general-purpose model with native computer use, tool search, and merged coding and reasoning in one frontier model.
Model Overview
Capabilities, design details, and architectural traits
GPT-5.4 - First General-Purpose Model with Native Computer Use
GPT-5.4 is the first OpenAI general-purpose model to ship native computer-use capabilities alongside reasoning, coding, and professional knowledge work in a single model. It merges the frontier coding strengths of GPT-5.3-Codex with GPT-5.2's reasoning base - a combination that previously required choosing between separate specialist models.
Four Specific Capabilities That Define This Release
| Trait | Detail |
|---|---|
| Native computer use | First general-purpose OpenAI model with built-in computer use; interprets screenshots and issues mouse/keyboard commands via the computer tool in the API; behavior is steerable and risk-configurable by developers |
| Tool search | Instead of loading all tool definitions into the prompt upfront, the model retrieves tool definitions on demand - documented to reduce token usage by 47% on MCP-heavy workloads while maintaining accuracy |
| Reasoning effort levels | Supports none, low, medium, high, and xhigh as explicitly documented reasoning effort settings, giving developers direct control over the compute-quality tradeoff per request |
| Token-efficient reasoning | Documented as the most token-efficient reasoning model in the GPT 5 series at release, solving problems with fewer tokens than GPT-5.2 |
The Codex Unification Decision
OpenAI explicitly named GPT-5.4 as a version-number jump - not a point release - because it unifies the Codex specialist and the mainline reasoning model into one. In Codex, GPT-5.4 replaces the previous pattern of routing between GPT-5.3-Codex (for code) and GPT-5.2 (for reasoning). It also introduces /fast mode in Codex, delivering up to 1.5x faster token velocity with no change to the underlying model weights.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | OpenAI | Anthropic | Anthropic |
| Release Date | March 5, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 1.1M Best Context Window | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $2.50 | $4 | $2 Best Input Pricing |
| Output Pricing | $15 | $20 | $10 Best Output Pricing |
| Modalities | |||
| Inputs | textimagefile | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 27.6 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from OpenAI
Other models by OpenAI
Top AI Models
Leading alternatives by intelligence score
