OpenAI Unveils 'GPT-4.1'... "Significant Improvement in Coding and Instruction Execution Abilities"

Provided via Developer API... Altman: "Focused on Real-World Usability"

ChatGPT developer OpenAI has unveiled a new artificial intelligence (AI) model, 'GPT-4.1.'

Reuters Yonhap News

On the 14th (local time), OpenAI announced that the performance of GPT-4.1, released on the same day, surpasses that of GPT-4o, launched last year, in almost every aspect.

Sam Altman, CEO of OpenAI, stated on the social networking service X (formerly Twitter), "GPT-4.1 excelled in coding, instruction execution, and handling long contexts," adding, "While the benchmark results are impressive, we focused on real-world usability, and developers' feedback has been very positive."

OpenAI explained that GPT-4.1's coding performance improved by 21% compared to GPT-4o and by 27% compared to GPT-4.5, which was introduced last February. They emphasized that the improvements in instruction compliance and long-context understanding make it more effective for operating AI agents.

According to OpenAI, GPT-4.1 can process up to 1 million tokens (basic units of words or sentences). This is about eight times the 128,000 tokens that GPT-4o can handle. The greater the number of tokens, the more data can be remembered at once, enabling uninterrupted understanding of long texts or complex conversations.

OpenAI also stated that the new model can reliably recognize given information while ignoring unnecessary elements compared to previous models. They added that such enhanced capabilities are essential for developing applications in legal and customer support fields.

Regarding cost-effectiveness, OpenAI explained, "GPT-4.1 is 26% cheaper than GPT-4o and successfully completed 54.6% of tasks in the coding benchmark (SWE-Bench) test, significantly outperforming GPT-4o's 33.2%."

GPT-4.1 will be initially available to developers through OpenAI's application programming interface (API). The integration schedule for general users into ChatGPT has not yet been announced.

Additionally, OpenAI introduced smaller models, GPT-4.1 mini and GPT-4.1 nano. GPT-4.1 mini offers faster computation speed and is relatively affordable, making it suitable for developers to experiment with, while GPT-4.1 nano is described by OpenAI as the smallest, fastest, and most affordable model among all OpenAI models released so far.