GPT-5.4 Outperforms Human Experts as AI Race Heats Up

A New Benchmark for Machine Intelligence

OpenAI's latest model, GPT-5.4 "Thinking," has crossed a threshold that the AI industry has long anticipated: on the company's GDPVal benchmark, which measures performance on real-world knowledge work across 44 occupations, the model matched or exceeded human professionals in 83% of comparisons. That's up from 70.9% for its predecessor, GPT-5.2, released just months earlier.

GDPVal isn't an abstract test. It evaluates AI on tasks drawn from the top nine industries contributing to U.S. GDP — generating sales presentations, building accounting spreadsheets, drafting urgent care schedules, and creating manufacturing diagrams. On spreadsheet modeling tasks typical of junior investment banking analysts, GPT-5.4 scored 87.3%, compared to 68.4% for GPT-5.2. Human raters preferred its presentation output 68% of the time.

Morgan Stanley Sounds the Alarm

The timing of GPT-5.4's release has amplified a stark warning from Morgan Stanley: a transformative AI leap is imminent, and most of the world isn't ready. In a report published in mid-March, the investment bank argued that unprecedented accumulation of compute power at American AI laboratories is driving capabilities faster than institutions can adapt.

The bank cited scaling laws suggesting that applying ten times the compute to large language model training effectively doubles a model's "intelligence" — and those laws are holding firm. Executives at major U.S. AI labs are reportedly telling investors to brace for progress that will "shock" them.

Morgan Stanley predicts that "Transformative AI" will become a powerful deflationary force, as AI tools replicate human work at a fraction of the cost. The bank noted that executives are already executing large-scale workforce reductions driven by AI efficiencies. The report also cited xAI co-founder Jimmy Ba, who suggested that recursive self-improvement loops — where AI autonomously upgrades its own capabilities — could emerge as early as the first half of 2027.

The Revenue Race Intensifies

The commercial stakes are enormous. OpenAI reached $25 billion in annualized revenue at the end of February 2026, growing 17% in just two months. Meanwhile, Anthropic hit $19 billion in annualized revenue in March 2026, up from $9 billion at the end of 2025 — a trajectory that, according to analysis by Epoch AI, could see Anthropic overtake OpenAI by mid-2026 or 2027.

The broader ecosystem is expanding rapidly. Yann LeCun's new startup AMI Labs raised $1.03 billion in what Sifted reported as Europe's largest-ever seed round, valued at $3.5 billion. AMI is betting on "world models" — AI that learns from physical reality, not just language — with applications in robotics, aerospace, and biomedical engineering.

Big Tech Doubles Down on Personalization

Google, meanwhile, is expanding its Personal Intelligence feature to all free-tier U.S. users. The tool, powered by Gemini, connects across a user's Google ecosystem — Gmail, Photos, Drive — to deliver highly personalized AI responses. Previously limited to paid subscribers, the rollout signals Google's strategy to embed AI deeply into everyday digital life, though the feature remains off by default and limited to personal accounts.

What Comes Next

The convergence of these developments — benchmark-shattering model performance, Wall Street warnings about workforce disruption, record startup funding, and aggressive consumer deployment — paints a picture of an industry accelerating beyond the capacity of regulators, employers, and workers to keep pace. Whether the world heeds Morgan Stanley's warning or not, the AI transformation is no longer a forecast. It's already underway.