OpenAI launches GPT-5.4: reasoning, coding, and computer use in one

"GPT-5.4 combines the coding capabilities of GPT-5.3-Codex with improvements in reasoning, knowledge work, and agentic workloads. On GDPval, a benchmark that assesses knowledge work across 44 professions in nine industries, GPT-5.4 achieves a score where it performs equally or better than human professionals in 83 percent of comparisons. GPT-5.2 stood at 70.9 percent."

"GPT-5.4 is the first general OpenAI model with native computer use capabilities. Like competing models, GPT-5.4 can control a computer via screenshots with mouse and keyboard commands without the need for external tools. On OSWorld-Verified, a benchmark focused on computer use tasks, GPT-5.4 achieves a success rate of 75.0 percent. That is above the human baseline of 72.4 percent."

"Developers working with the API also benefit from tool search. Instead of always loading all tool definitions in context, the model searches for the required tool itself at the right moment. In a test with 250 tasks across 36 MCP servers, this approach reduced token usage by 47 percent with equal accuracy."

GPT-5.4 represents OpenAI's latest advancement, combining enhanced reasoning capabilities with native computer control features. The model achieves 83% performance equality with human professionals on GDPval, a knowledge work benchmark spanning 44 professions across nine industries, compared to GPT-5.2's 70.9%. GPT-5.4 introduces native computer use capabilities, enabling mouse and keyboard control via screenshots, achieving 75% success on OSWorld-Verified benchmarks, exceeding the human baseline of 72.4%. The model integrates coding capabilities from GPT-5.3-Codex with improved agentic workloads. A Pro version offers enhanced performance for complex tasks. Tool search functionality reduces API token usage by 47% while maintaining accuracy, improving cost efficiency and speed for developers.

Read at Techzine Global