OpenAI on Monday evening introduced the GPT-4.1 series, featuring models like GPT-4.1 nano, with what it said were notable advancements in coding, long context and instruction following.
These models promise major improvements over their predecessors, such as better performance on coding tasks and the capability to manage up to a million tokens of context, a significant leap from previous limits.
According to OpenAI, GPT-4.1 significantly outperforms models like GPT-4o and the soon-to-be-deprecated GPT-4.5 Preview, with superior scores on several benchmarks.
GPT-4.5-Preview was the company’s most expensive model and didn’t receive favorable feedback from the developers.
OpenAI said it would be turned off in three months, on July 14, as the new model would offer “similar performance on many key capabilities at much lower cost and latency.”
Notably, in software engineering tasks, GPT-4.1 achieves a 54.6% completion rate on the SWE-bench Verified, representing substantial progress over previous models.
The standout feature of these models is their extraordinary long-context processing.
GPT-4.1 models can handle the equivalent of more than eight React codebases within one window and find relevant snippets regardless of their position.
Interestingly, while GPT-4.1 is exclusive to the API, many improvements have gradually unfolded in ChatGPT’s latest version with plans to incorporate even more features in future iterations.
Meanwhile, GPT-4.1 models promise high efficiency and reduced costs – notably, it is 26% cheaper than its predecessors for median queries, with the GPT-4.1 nano model setting a new benchmark for affordability and speed.
Planning your financial journey can be daunting but it doesn't have to be. Fire Fast by Dzambhala helps you understand and plan effectively.
Join the vibrant privacy-ensured Dzambhala community on
Want to give feedback on this story? Write to us.