tech
March 5, 2026
OpenAI launches GPT-5.4 with Pro and Thinking versions
GPT-5.4 is billed as "our most capable and efficient frontier model for professional work."
TL;DR
- OpenAI released GPT-5.4, a new foundation model for professional work, available in standard, reasoning, and high-performance versions.
- The API version features a 1 million token context window, the largest offered by OpenAI.
- GPT-5.4 shows improved token efficiency, requiring fewer tokens to solve the same problems as its predecessor.
- The model achieved record scores on computer use benchmarks (OSWorld-Verified, WebArena Verified) and OpenAI's GDPval test for knowledge work.
- It also led Mercor's APEX-Agents benchmark for professional skills in law and finance.
- GPT-5.4 is 33% less likely to make errors in individual claims and 18% less likely to have errors in overall responses compared to GPT 5.2.
- A new 'Tool Search' system for API tool calling reduces token consumption and improves efficiency.
- New safety evaluations indicate that deception is less likely in GPT-5.4 Thinking, suggesting CoT monitoring remains effective.
Continue reading the original article