How GPT‑5.4 Redefines Computer Use and Agentic Workflows

OpenAI has announced the release of GPT‑5.4, its most advanced AI model to date, designed to enhance professional workflows across reasoning, coding, computer use, and agentic tasks. Building on the capabilities of GPT‑5.2 and GPT‑5.3 Codex, this new model introduces significant improvements in efficiency, accuracy, and usability.

Key Features

• Two Versions:

  • GPT‑5.4 Thinking – optimized for reasoning and structured problem-solving.
  • GPT‑5.4 Pro – optimized for maximum performance in complex, long-horizon tasks.

• Expanded Context Window: Up to 1 million tokens in Codex (experimental), enabling deeper document and code analysis.

• Efficiency Gains: Uses fewer tokens compared to GPT‑5.2, reducing costs and latency.

Performance Improvements

• Reasoning & Knowledge Work:
  GPT‑5.4 achieved an 83% win rate in GDPval benchmark tasks, outperforming GPT‑5.2’s 70.9%. Human evaluators preferred GPT‑5.4’s presentations nearly 68% of the time.

• Coding:
  Enhanced debugging and iteration speed with a new /fast mode, delivering up to 1.5x faster token velocity. Integration with Playwright Interactive allows live app testing and debugging.

• Computer Use & Vision:
  GPT‑5.4 introduces native computer-use capabilities, issuing mouse and keyboard commands based on screenshots. It scored 75% on the OSWorld-Verified benchmark, surpassing human performance. Visual perception has also improved, supporting 10.24M pixel images.

• Web Research & Tool Use:
  With tool search functionality, GPT‑5.4 reduces token usage by nearly half in tool-heavy workflows. On the BrowseComp benchmark, GPT‑5.4 Pro achieved 89.3% success, demonstrating stronger persistence in multi-step research tasks.

Industry Applications

• Legal: Scored 91% on BigLaw Bench, excelling in contract analysis.
• Finance: Outperformed GPT‑5.2 in investment banking modeling tasks (87.3% vs. 68.4%).
• Enterprise Tools: Launch of a new ChatGPT for Excel add-in alongside GPT‑5.4.
• Developer Workflows: Reduced hallucinations by 33%, improving reliability in agentic tasks.

Safety & Pricing

GPT‑5.4 is classified as High cyber capability under OpenAI’s Preparedness Framework, with enhanced safeguards against misuse.

• Pricing (API):
  • GPT‑5.4: $2.50/M input tokens, $15/M output tokens
  • GPT‑5.4 Pro: $30/M input tokens, $180/M output tokens
  • Batch/Flex pricing available at half rate; Priority processing at double rate