OpenAI Releases GPT-4.1 for the API

OpenAI has launched the GPT-4.1 series via API, featuring greatly improved coding, instruction-following, and long-context understanding capabilities. This series consists of GPT-4.1, GPT-4.1 mini, and the first-ever nano model, GPT-4.1 nano. All models support a context window of up to 1 million tokens, and the knowledge cutoff has been updated to June 2024.

Key performance metrics are as follows:

• Coding: Achieved 54.6% on SWE-bench Verified, an improvement of 21.4 points over GPT-4o and 26.6 points over GPT-4.5.
• Instruction-following: Scored 38.3% on the Scale MultiChallenge benchmark, up 10.5 points from GPT-4o.
• Long-context understanding: Achieved 72.0% on Video-MME, setting a new SOTA (State-of-the-art).

As for model-specific features, GPT-4.1 mini shows intelligence on par with or higher than GPT-4o while cutting latency in half and reducing cost by 83%. GPT-4.1 nano is the fastest and cheapest model, supporting a context of 1 million tokens and optimized for tasks such as classification or autocompletion.

The enhanced reliability and context understanding are highly effective for implementing AI agents. Meanwhile, OpenAI plans to sunset GPT-4.5 Preview on July 14, 2025, to encourage the transition to GPT-4.1.