GPT-5.4 is OpenAI's most advanced language model, released March 5, 2026. It scored 83% on the GDPVal benchmark — at or above human expert level across 44 professional occupations including medicine, law, finance, and software engineering. Available in two variants: GPT-5.4 Pro for speed and everyday tasks, and GPT-5.4 Thinking for complex multi-step reasoning requiring deeper analysis.

How does GPT-5.4 Pro differ from Thinking?

GPT-5.4 Pro is optimized for speed and everyday professional tasks — fast responses, code generation, writing, and analysis. GPT-5.4 Thinking uses extended reasoning chains, spending more time thinking through complex problems before responding. Thinking is better for math, logic, and scientific reasoning, while Pro handles rapid iteration and high-volume tasks efficiently.

What does the GDPVal benchmark measure?

GDPVal measures an AI model performance across 44 professional occupations based on actual economic value — tasks like medical diagnosis, legal analysis, financial modeling, and software engineering. A score above 80% indicates performance at or above human expert level. GPT-5.4 scored 83%, making it the first publicly available AI to achieve this milestone across such a broad range of professions.

Can GPT-5.4 use a computer autonomously?

Yes, GPT-5.4 achieved 75% on the OSWorld-Verified computer use benchmark, surpassing average human performance of 72.4%. This means GPT-5.4 can navigate desktop interfaces, use applications, browse the web, and complete multi-step computer tasks with a higher success rate than most human users — a significant milestone for autonomous AI agents in professional settings.

How much does GPT-5.4 cost?

GPT-5.4 is available through OpenAI API with tiered pricing for Pro and Thinking variants. The Pro version is priced for high-volume professional use, while Thinking is more expensive due to extended compute required for reasoning. ChatGPT Plus and Pro subscribers get access to both models. Exact API pricing varies by token count and is listed on OpenAI pricing page at platform.openai.com.

Source: OpenAI

/// Tech Trend 2026

GPT-5.4: OpenAI's Most Powerful AI Surpasses Human Experts

Published: March 5, 2026

OpenAI releases GPT-5.4 — the most advanced frontier model for professional work. Scoring 83% on GDPVal and 75% on OSWorld, this is a leap that redefines human productivity.

/// Key Metrics

Performance That Breaks Boundaries

GDPVal Score — at/above human expert level

OSWorld-Verified — surpassing humans (72.4%)

$0B+

OpenAI annualized revenue milestone

Professional occupations evaluated in GDPVal

/// Variant Comparison

GPT-5.4 Pro vs GPT-5.4 Thinking

GPT-5.4 Pro

Speed & Efficiency

Optimized for everyday professional tasks
Faster responses with high accuracy
Ideal for large-scale API integration
Lower inference cost than GPT-5.4 Thinking
Multimodal: text, images, and code

GPT-5.4 Thinking

Deep Reasoning

Extended reasoning chains for complex problems
Top GDPVal performance at 83%
Excellence in math, coding, and science
Legal and medical analysis at expert level
75% OSWorld — autonomous computer use

/// GDPVal by Profession

GPT-5.4 vs Human Experts Across Professions

Software Engineer

GPT-5.4: 91%Human: 84%

Lawyer

GPT-5.4: 85%Human: 87%

Physician

GPT-5.4: 82%Human: 89%

Accountant

GPT-5.4: 88%Human: 81%

Financial Analyst

GPT-5.4: 86%Human: 83%

Public Health Expert

GPT-5.4: 79%Human: 82%

Data from GDPVal benchmark — OpenAI, March 2026. Results are representative samples.

/// Industry Impact

Which Industries Is GPT-5.4 Transforming?

Technology & Software

Automated coding reaches 91% success on real-world tasks. Engineering team productivity increases 3–5x per internal reports.

91%

Legal & Compliance

Contract drafting, case research, and legal risk analysis at or above mid-level attorney performance.

85%

Healthcare

Diagnostic support, medical imaging analysis, and clinical research synthesis approaching specialist physician accuracy.

82%

Finance & Investment

Portfolio analysis, risk modeling, and market forecasting being integrated into major fund workflows.

86%

Morgan Stanley — H1 2026 Warning

Morgan Stanley warns that a transformative AI leap is imminent in H1 2026, as models like GPT-5.4 begin displacing knowledge work at meaningful scale. OpenAI has surpassed $25B in annualized revenue, signaling accelerating enterprise adoption.

/// OpenAI Model History

The OpenAI Journey 2023–2026

Mar 2023GPT-4

OpenAI's first multimodal model, setting a new benchmark for LLMs.

Nov 2023GPT-4 Turbo

128K context window, 3x cost reduction vs GPT-4.

May 2024GPT-4o

Real-time responses with audio, vision, and text in one unified model.

Sep 2024o1

First reasoning model with chain-of-thought, achieving PhD-level math scores.

May 2025GPT-5

Major leap in multi-task performance, surpassing GPT-4o on all major benchmarks.

Mar 5, 2026GPT-5.4LATEST

Achieves 83% GDPVal and 75% OSWorld. Most advanced model for professional work.

/// Visual Documentation

Visual Evidence

GPT-5.4 benchmark comparison with competing models

GPT-5.4 impact on AI-driven knowledge worker productivity

/// Frequently Asked Questions

Questions Answered

/// References

Sources

Introducing GPT-5.4 — OpenAI
OpenAI launches GPT-5.4 with Pro and Thinking versions — TechCrunch
Morgan Stanley warns AI breakthrough coming 2026 — Fortune

Read more related AI trends: DeepSeek V4 — China's AI rival and OpenAI valuation 2026.

▸ GPT-5.4 Pro costs $200/month — about one restaurant meal per week, but it can replace an entire analytics team's workload.

▸ With 40% better coding ability than GPT-4o, Vietnamese developers can significantly boost productivity using GPT-5.4.

GPT-5.4 — Released: March 5, 2026 · OpenAI · Technology · SEOFactory