OpenAI has released three new models—GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano—in its API, featuring improved coding, instruction following, and long-context capabilities compared to GPT-4o.

  • GPT-4.1 scores 54.6% on SWE-bench Verified, a 21.4% absolute improvement over GPT-4o, and sets a new state-of-the-art on Video-MME for multimodal long-context understanding.
  • GPT-4.1 mini reduces latency by nearly half and cost by 83% while matching or exceeding GPT-4o in intelligence evals.
  • GPT-4.1 nano is the fastest and cheapest model, scoring 80.1% on MMLU and supporting a 1 million token context window.
  • All models support up to 1 million tokens of context and feature a knowledge cutoff of June 2024.

These improvements make the GPT-4.1 family more effective for building agents and handling complex tasks like software engineering and document analysis.