OpenAI launched its new flagship model GPT-5.4 on Thursday while facing a significant user backlash. The “QuitGPT” movement, an estimated 2.5 million users taking action against the company, emerged after OpenAI secured a controversial defense contract that rival Anthropic refused. The new model boasts a 1-million-token context window, stronger reasoning, and improved efficiency, with enterprise users poised to gain the most from its faster, more token-efficient agentic capabilities.
OpenAI began rolling out GPT-5.4 as it contends with a public relations crisis. An estimated 2.5 million users have canceled subscriptions or shared the “QuitGPT” boycott following the company’s deal with the U.S. Department of Defense.
Anthropic publicly walked away from the same Pentagon contract, earning scorn from government officials. The sticking point was the DoD’s refusal to include language prohibiting autonomous weapons and mass surveillance of U.S. citizens.
The new model consolidates reasoning, coding, and agentic capabilities into a single release just two days after GPT-5.3 was introduced. It features a million-token context window and a mid-response steering feature that lets users redirect the model while it’s thinking.
On the GDPval benchmark, GPT-5.4 matches or beats professionals in 83.0% of comparisons, up from 70.9% for GPT-5.2. Its computer use success rate on OSWorld-Verified hit 75.0%, clearing the human baseline of 72.4%.
OpenAI stated, “GPT‑5.4 is our most token-efficient reasoning model yet, using significantly fewer tokens to solve problems when compared to GPT‑5.2.” This efficiency is crucial for API users charged per token.
The clearest beneficiaries are enterprise users performing document-heavy work. Legal research firm Harvey reported a 91% score on its BigLaw Bench evaluation, while agent operator Mainstay saw sessions run three times faster using 70% fewer tokens.
