OpenAI GPT-4.1

Launched April 14, 2025 in three sizes (GPT-4.1, mini, nano), GPT-4.1 was OpenAI's developer-targeted refinement of the GPT-4 line: a full 1-million-token context window, 32K output tokens, and training tuned for real-world software engineering — 54.6% on SWE-bench Verified versus 33.2% for GPT-4o. It reliably attends across the full context length, follows instructions more precisely (38.3% on Scale's MultiChallenge), and handles images and video better than GPT-4o. The nano variant became OpenAI's smallest, fastest, cheapest model, and GPT-4.1 mini replaced GPT-4o mini as the free-tier fallback in ChatGPT.\n\nLimitations: still weaker than contemporaneous reasoning models (o3, Gemini 2.5 Pro, Claude 3.7) on hard multi-step reasoning. Knowledge cutoff is mid-2024. It never shipped a dedicated reasoning mode — it is a fast non-reasoning workhorse, not a deliberative thinker.

OpenAI-reported: 54.6% SWE-bench Verified, 38.3% Scale MultiChallenge