OpenAI GPT-4o and GPT-4.1 Review — Omni, Agentic, and Still Catching Up

GPT-4o launched May 2024 as OpenAI's first natively multimodal model — unified text, audio, image, and video in a single network. GPT-4.1 followed April 2025 with 1M-token context, 54.6% SWE-bench coding, and API-first agentic design. We review both models: architecture, benchmarks, pricing, the Studio Ghibli copyright moment, the sycophancy rollback, and where OpenAI's flagship models stand against Claude, Gemini, and DeepSeek.