HIGHLIGHTS
GPT-4.1 scores 54.6% on SWE-bench Verified, outperforming GPT-4o by 21.4%.
Models offer up to one million tokens of context, ideal for complex tasks.
GPT-4.1 mini and nano offer faster performance with reduced costs.
OpenAI has introduced three new API models, including GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, with enhancements in coding, instruction following, and long-context comprehension. The GPT 4o series successor offers fast processing, better accuracy, and a larger context window of up to one million tokens.
According to the company, the GPT 4.1 has achieved 4.6% on SWE-bench Verified, a benchmark for software engineering tasks, outperforming GPT-4o by 21.4%. The model also scores 38.3% on the Scale’s MultiChallenge benchmark, a test of instruction-following ability, marking a 10.5% improvement. This means the GPT 4.1 is more reliable in generating code, following instructions, and handling detailed tasks.
Taking to a blog post, the company stated that GPT-4.1 models can process up to one million tokens of context, nearly eight times the size of the entire React codebase. It means that the models offer improved performance in retrieving and understanding information scattered across long documents , making them suitable for complex tasks including legal analysis and multi document review.
Also read: Google Pixel 10 Pro vs Pixel 9 Pro: Price, camera, battery, design and other upgrades you can expect
Interestingly, the GPT 4.1 series, as per the company, offers improved performance at lower costs, with GPT-4.1 mini offering high performance with reduced latency and 83% lower costs compared to the predecessors. On the other hand, the company claimed that GPT-4.1 nano is the fastest model in the series.
Additionally, the company has announced plans to retire the GPT-4.5 Preview by July 14, 2025, as the new models offer similar or better performance at lower costs. These models are available via OpenAI’s API, with a pricing structure designed to be more affordable for developers.
Follow Us
Ashish Singh
Ashish Singh is the Chief Copy Editor at Digit. He’s been wrangling tech jargon since 2020 (Times Internet, Jagran English ’22). When not policing commas, he’s likely fueling his gadget habit with coffee, strategising his next virtual race, or plotting a road trip to test the latest in-car tech. He speaks fluent Geek. View Full Profile
