OpenAI Launches GPT-5.4 Model
OpenAI just released GPT-5.4, its most powerful AI model yet, designed for enterprise and technical tasks. The model features a massive 1-million-token context window and a new “native computer use” capability, allowing it to autonomously operate across a user's apps and files. The release is seen as a direct challenge to competitors, especially with Anthropic now facing U.S. government sanctions.
The new model arrives in two versions: GPT-5.4 Thinking, which is now the default for Plus, Team, and Pro subscribers, and a more powerful GPT-5.4 Pro for enterprise clients. OpenAI will phase out the older GPT-5.2 Thinking model, making it available as a legacy option until its discontinuation on June 5, 2026. On key benchmarks, GPT-5.4's performance now exceeds human levels in certain tasks. It achieved a 75% success rate on the OSWorld Verified benchmark for navigating desktop environments, surpassing the human average of 72.4%. For professional knowledge work, it scored 83% on the GDPval benchmark, a significant leap from the 70.9% achieved by GPT-5.2. Under the hood, this is the first OpenAI model to integrate the advanced coding capabilities of GPT-5.3 Codex with reasoning and agentic workflows. For developers, a new API feature called "Tool Search" reduces token usage by 47% in tests by retrieving tool definitions only when necessary. The model has also received a "High Capability" rating for cybersecurity, a first for one of OpenAI's general models. The enterprise push extends beyond the model itself, with OpenAI also launching a beta of ChatGPT for Excel and new financial data integrations. The company's internal tests showed GPT-5.4 scoring 87.3% on spreadsheet modeling tasks that a junior investment banking analyst might perform, compared to 68.4% for GPT-5.2. The U.S. government's sanctions against rival Anthropic stem from a contract dispute with the Department of Defense. Anthropic refused to remove contractual "red lines" that would prevent its Claude AI models from being used for mass domestic surveillance or in fully autonomous weapons systems without human oversight. As a result, the DoD designated Anthropic a "supply-chain risk to national security." The directive not only ends Anthropic's government contracts but also bars any company doing business with the U.S. military from conducting any commercial activity with Anthropic, forcing them to certify their compliance.
