GPT-5.4
FrontierOpenAI•Released on 2026-03-05
OpenAI's most capable and efficient frontier model for professional work. Combines industry-leading coding with native computer use, 1M+ context window, and improved reasoning. First GPT model to beat human performance on desktop navigation tasks.
91
Overall Score
Core Specs
1050K
Context Window
128K
Max Output
✓
Reasoning
✗
Open Source
Multimodal Support
textimage
Scenario Scores
User Feedback Highlights
Based on community feedback. Hover to see original reviews.
+ 33% fewer hallucinations vs GPT-5.2+ Tool search reduces tokens by 47%− 2x pricing above 272K tokens+ Native computer use capability− Health benchmarks slightly worse than 5.2+ 1M+ context window (largest in GPT lineup)− Some users report benchmark-optimized feel+ Half the price of Claude Opus 4.6− 24% longer average responses (more output tokens)
Sentiment:👍 58%😐 28%👎 14%
Pros & Cons
Pros
- +1M+ context window (largest in GPT lineup)
- +Native computer use capability
- +33% fewer hallucinations vs GPT-5.2
- +Tool search reduces tokens by 47%
- +Half the price of Claude Opus 4.6
Cons
- −2x pricing above 272K tokens
- −24% longer average responses (more output tokens)
- −Health benchmarks slightly worse than 5.2
- −Some users report benchmark-optimized feel
Reliability
Pricing
Input (per 1M tokens)$2.50
Output (per 1M tokens)$15.00
Subscription$20/month
Updated on 2026-03-06
Tools Supporting This Model
Compare with Others
Benchmarks
sweBenchPro57.7%
gdpval83%
osWorld75%
mmmuPro81.2%
humanPerformance72.4%