GPT-5.4 Pro

OpenAICurrentReleased 2026-03-05
Overall Score
92#1 of 104
Arena Elo
1484
Categories Ranked
8of 8
Price (1M tokens)
$30 in / $180 out
Speed
74tok/s
Context
1.05M
ProprietaryReasoning
Confidence
pro

Ranking Distribution

Category rank across 8 benchmark categories (lower is better)

Category Performance

Scores across all benchmark categories (0-100 scale)

Category Breakdown

Agentic

#1
88.9/ 100
Weight: 22%3 benchmarks
Terminal-Bench 2.0BrowseCompOSWorld-Verified

Coding

#1
88.3/ 100
Weight: 20%4 benchmarks
SWE-bench VerifiedLiveCodeBenchSWE-bench ProSWE-Rebench

Reasoning

#1
95.7/ 100
Weight: 17%4 benchmarks
MuSRLongBench v2MRCRv2ARC-AGI-2

Knowledge

#2
84.9/ 100
Weight: 12%7 benchmarks
GPQASuperGPQAMMLU-ProHLEFrontierScienceSimpleQA

Math

#1
98.5/ 100
Weight: 5%8 benchmarks
AIME 2025BRUMO 2025MATH-500

Multilingual

#1
95.7/ 100
Weight: 7%2 benchmarks
MGSMMMLU-ProX

Multimodal

#4
94.9/ 100
Weight: 12%2 benchmarks
MMMU-ProOfficeQA Pro

Inst. Following

#1
96.9/ 100
Weight: 5%1 benchmark
IFEval

Chatbot Arena Performance

Text Overall1484CI: ±7.47,160 votes
Coding1533CI: ±14.91,596 votes
Math1517CI: ±28.3428 votes
Instruction Following1488CI: ±13.51,988 votes
Creative Writing1461CI: ±19.41,004 votes
Multi-turn1502CI: ±16.11,361 votes
Hard Prompts1508CI: ±9.54,032 votes
Hard Prompts (English)1510CI: ±14.11,801 votes
Longer Query1492CI: ±13.22,044 votes

Benchmark Details

Compare This Model

See how GPT-5.4 Pro stacks up against similar models

GPT-5.4 Family

Other variants in the same model family