OpenAI adds GPT‑5.4
OpenAI updated ChatGPT to include GPT‑5.3 and a GPT‑5.4 “Thinking” variant, which the company describes as its most capable reasoning model inside ChatGPT. The release notes call out improved performance on spreadsheets, code-generation, document and image understanding, and tool use — positioning the model explicitly for multi‑step, output‑oriented tasks. (help.openai.com)
OpenAI has added GPT‑5.4 Thinking to ChatGPT and made GPT‑5.3 Instant the default model for logged-in users. (help.openai.com) OpenAI said GPT‑5.3 Instant is rolling out to all ChatGPT users, while GPT‑5.4 Thinking is the company’s “most capable reasoning model” inside ChatGPT for harder tasks. The change appears in OpenAI’s ChatGPT help documentation updated in early April 2026. (help.openai.com) The company said GPT‑5.3 Instant is tuned for everyday work and learning, including info-seeking questions, how-to requests, technical writing, and translation. GPT‑5.4 Thinking is positioned for spreadsheet creation and editing, frontend code, slideshow creation, hard math, document understanding, image understanding, tool use, and web research that combines many sources. (help.openai.com) ChatGPT now uses an auto-switching system under the “Instant” setting, which can route more complex requests from GPT‑5.3 Instant to GPT‑5.4 Thinking. Users can also pick “Thinking” directly from the model picker for deeper reasoning. (help.openai.com) A reasoning model is built to spend more time working through steps before answering, instead of replying as quickly as possible. OpenAI said GPT‑5.4 Thinking can show a short plan before it starts, accept new instructions while it is still working, and keep track of prior steps longer without timing out. (openai.com) OpenAI released GPT‑5.4 on March 5, 2026, and said the model combines recent work in reasoning, coding, and “agentic workflows,” meaning systems that can use tools and carry out multi-step jobs. In the application programming interface and Codex, OpenAI said GPT‑5.4 also supports up to 1 million tokens of context and native computer-use capabilities. (openai.com) The company framed the update around office-style output rather than chat alone. OpenAI said GPT‑5.4 improves work across spreadsheets, presentations, documents, software tools, and connectors, with the goal of producing finished results with less back-and-forth. (openai.com) OpenAI also published benchmark results alongside the launch, saying GPT‑5.4 outperformed GPT‑5.2 on tests including GDPval, SWE‑Bench Pro, OSWorld‑Verified, Toolathlon, and BrowseComp. On GDPval, which OpenAI describes as a knowledge-work benchmark across 44 occupations, GPT‑5.4 scored 83.0 percent versus 70.9 percent for GPT‑5.2. (openai.com) The release lands after a fast model reshuffle inside ChatGPT in March 2026. OpenAI’s model release notes say GPT‑5.1 models were retired on March 11, 2026, and existing conversations were moved to GPT‑5.3 Instant, GPT‑5.4 Thinking, or GPT‑5.4 Pro. (help.openai.com) OpenAI’s help center says GPT‑4o, GPT‑4.1, GPT‑4.1 mini, OpenAI o4-mini, and GPT‑5 Instant and Thinking were retired from ChatGPT on February 13, 2026. The result is a simpler split inside ChatGPT: a fast default model for most prompts, and a slower reasoning model for work that needs more steps. (help.openai.com)
