OPENAI GPT-5.4 REACHES NEW BENCHMARKS AS MODEL WARS ACCELERATE

OpenAI’s GPT-5.4, released last month, continues to set new standards for AI capability. The model now scores 83% on GDPVal, a benchmark testing professional-quality work across 44 jobs — a 12-point jump from its predecessor’s 70.9% just months earlier. Morgan Stanley analysts call the April-June window a “nonlinear increase” in AI capability, warning that most organizations are unprepared for the pace of change.

Meanwhile, the competitive landscape intensifies. Google announced Gemma 4 on April 2, an open-weights model built from the same technology as Gemini 3. Microsoft unveiled three in-house AI models on the same day — directly competing with OpenAI, its longtime partner. The Verge described Microsoft’s move as “shivving OpenAI” with homegrown models for speech and images.

The implications are stark: enterprises that delayed AI adoption are now racing to catch up. A Morgan Stanley report warns that the “hard takeoff” is already underway. Elon Musk, speaking at the 2026 Abundance Summit, said: “I go to sleep, there’s some massive AI breakthrough, and when I wake up, there’s another one.”

Sources:

• The Verge: OpenAI’s new GPT-5.4 model
• Morgan Stanley: AI Leap analysis
• Google DeepMind: Gemma 4 announcement
• Microsoft: Three new foundational models