OpenAI logo

GPT-5.4

OpenAI · Released Mar 5, 2026

ProprietarycodingreasoningvisionCompare this model

$2.5

Input price / MTok

#21 of 27

$15

Output price / MTok

#22 of 27

1.0M

Context window

#9 of 28

165

Tokens / sec

#6 of 21

57.7

SWE-bench Pro

#8 of 9

75

OSWorld

#6 of 8

36.6

MRCR

#5 of 5

Summary

GPT-5.4 is a proprietary language model from OpenAI, released on Mar 5, 2026.

At $2.5 input and $15 output per million tokens, it ranks #21 of 27 for input price, placing it mid-range on input cost.

Its context window of 1.0M tokens ranks #9 of 28 for context depth in this dataset.

It targets coding, reasoning, vision workloads.

Throughput is measured at 165 tokens per second (ranked #6 of 21 for throughput).

75% computer-use benchmark; 1M+ context (922K input + 128K output).

Benchmarks

BenchmarkScoreMax%RankSource
SWE-bench Pro57.710057.7%#8/9OpenAI (self-reported in GPT-5.4 launch,...
OSWorld7510075.0%#6/8OpenAI (self-reported in GPT-5.4 launch,...
MRCR36.610036.6%#5/5OpenAI (self-reported in GPT-5.4 launch,...

Similar Models

Compare all 6 models

Models with the most shared benchmark coverage and closest scores to GPT-5.4.

GPT-5.4
OpenAIGPT-5.5
GoogleGemini 3.5 Flash
AnthropicClaude Sonnet 4.6
AnthropicClaude Sonnet 5
AnthropicClaude Opus 4.8
SWE-bench Pro#8 of 9
Professional software engineering tasks
OSWorld#6 of 8
Computer-use and GUI automation
MRCR#5 of 5
Long-context retrieval and recall

Compare GPT-5.4 side by side

Pre-filled with 5 similar models

Open comparison

Speed source: https://artificialanalysis.ai/models/gpt-5-4