Skip to main content
← Back to sources

OpenAI Launches GPT-5.4 with Pro and Thinking Versions for Professional Work

Published 2026-03-05Ingested 2026-03-06Foundation ModelsHigh⭐ Timeline Candidate

Summary

OpenAI has released GPT-5.4, described as its most capable and efficient frontier model for professional work, available in both Pro and Thinking versions. The model achieved record scores across multiple benchmarks, including computer use benchmarks OSWorld-Verified and WebArena Verified, as well as an 83 percent score on OpenAI's GDPval test for knowledge work tasks. GPT-5.4 also took the lead on Mercor's APEX-Agents benchmark, which tests professional skills in law and finance. Mercor CEO Br

Alignment: Reinforces current position
Related Positions: multi-model-multi-vendor.md, agentic-workflows.md, enterprise-ai-delivery.md, ai-assisted-development-tooling.md
Related Partnerships: microsoft-github.md
openaigpt-5-4foundation-modelsagentic-benchmarkscomputer-useprofessional-workapi-updatesknowledge-workmulti-model-strategyfrontier-models
OpenAI Launches GPT-5.4 with Pro and Thinking Versions for Professional Work — Intelligence — Agentic Developer Tools Radar · Signal