OpenAI has now rolled out GPT-5.4, its latest frontier large language model family, which the firm says features significant advancements in reasoning, coding, multimodal understanding, long-context handling, and native computer-use capabilities. The release includes two primary variants available in ChatGPT: GPT-5.4 Thinking for Plus, Team, and Pro subscribers, and GPT-5.4 Pro in ChatGPT, with full rollout across the API and Codex. This positions OpenAI closer to fully autonomous AI agents capable of end-to-end task execution, potentially transforming enterprise workflows in coding, analytics, finance, and customer service.
“Today, we’re releasing GPT‑5.4 in ChatGPT (as GPT‑5.4 Thinking), the API, and Codex. It’s our most capable and efficient frontier model for professional work. We’re also releasing GPT‑5.4 Pro in ChatGPT and the API, for people who want maximum performance on complex tasks,” the company announced in a blog post. “In ChatGPT, GPT‑5.4 Thinking can now provide an upfront plan of its thinking, so you can adjust course mid-response while it’s working, and arrive at a final output that’s more closely aligned with what you need without additional turns. GPT‑5.4 Thinking also improves deep web research, particularly for highly specific queries, while better maintaining context for questions that require longer thinking. Together, these improvements mean higher-quality answers that arrive faster and stay relevant to the task at hand.”
GPT-5.4 replaces the previous GPT-5.2 Thinking model (set for deprecation in three months) and builds on recent updates to the Instant model focused on more natural conversation. The new family focuses on agentic workflows (enabling AI to interact with software, browsers, and tools autonomously) while delivering measurable improvements in factual accuracy, token efficiency, and other areas. OpenAI described GPT-5.4 as its “most capable and efficient frontier model for professional work” with individual claims 33% less likely to be false and overall responses 18% less prone to errors compared to GPT-5.2, according to internal benchmarks.
To provide some context, the company’s frontier model lineage has accelerated since the GPT-5 series debuted, with GPT-5.2 released in December as a “code red” response to competitive pressure from Google Gemini and Anthropic’s Claude. The latest iteration arrives amid ongoing rivalry and heightened public scrutiny of AI safety and military applications. The company has positioned GPT-5.4 as a step toward agentic AI systems capable of executing multi-step tasks across applications. It introduces native computer-use features—allowing the model to issue mouse/keyboard commands, interact with screenshots, and operate software via libraries like Playwright.
GPT-5.4 Thinking provides a visible plan-of-action outline for complex queries, allowing users to intervene mid-response to adjust direction without restarting. This feature is live on web and Android, with iOS rollout coming soon. It can also autonomously control computers, verify actions, and iterate in build-run-verify-fix loops. It achieved a record 75% on the OSWorld-Verified benchmark (surpassing human testers at 72.4%) and supports long-running, multi-step agent trajectories. In addition to this, the 1-million-token context window enables analysis of entire codebases, large document sets, or extended histories. Image perception supports inputs over 10 million pixels without compression, improving detail retention. Substantial gains in spreadsheet analysis (87.3% benchmark score, up from 68.4% in GPT-5.2) as well.
The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure →
