OpenAI Releases GPT-5.4
OpenAI's GPT-5.4 release brings major upgrades for professional workflows and analytics. This includes improved code generation, creative writing, and longer context windows, with a new 'Thinking' mode allowing real-time direction. Consider experimenting with GPT-5.4 for auto-generating SQL queries, basic Python scripts, or even dashboard mockups for your portfolio.
GPT-5.4, launched on March 5, 2026, integrates coding skills from GPT-5.3 Codex with improved reasoning and accuracy. OpenAI reports a 33% reduction in false claims and an 18% reduction in errors compared to GPT-5.2. It's available through ChatGPT, the API, and Codex, targeting both casual and enterprise-level users. GPT-5.4 achieves state-of-the-art performance on the GDPval benchmark, matching or exceeding industry professionals in 83% of comparisons across 44 occupations; GPT-5.2 only achieved 70.9%. The model particularly excels in spreadsheet and presentation-heavy tasks, with an 87.3% score on internal spreadsheet modeling benchmarks compared to GPT-5.2's 68.4%. Enterprise customers can use GPT-5.4 with the ChatGPT for Excel add-in for improved spreadsheet creation and analysis. A key feature is native computer use, enabling the model to operate desktop applications autonomously. It can navigate UIs, click elements, fill forms, and move between windows, achieving a 75% success rate on the OSWorld-Verified benchmark, surpassing the 72.4% human baseline. GPT-5.4 also supports a 1,050,000 token context window, allowing it to analyze entire codebases or large document collections in a single request. GPT-5.4 is offered in standard, Thinking, and Pro variants. The Thinking variant can provide an upfront plan of its reasoning, allowing users to adjust the course mid-response. While GPT-5.4 boasts improved token efficiency, using 47% fewer tokens on complex tasks, its input cost is higher at $2.50 per million tokens compared to GPT-5.2's $1.75. GPT-5.2 Thinking will be retired on June 5, 2026.
