Tech Desk 
Apr 15, 2025
In a bid to cater to the overwhelming demand for artificial intelligence's (AI) assistance in programming and coding, OpenAI has introduced a new assortment of AI models, including GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, all specially designed for tasks involving software coding and instruction-following.
The new family of GPT-4.1 models can be accessed from OpenAI’s API.
Containing an enormous context window of 1 million tokens, meaning the ability to process around 750,000 words, they're developed to handle extensive tasks, including end-to-end app development.
The launch of OpenAI's latest AI models coincides with intense competition from arch-rivals such as Google's Gemini 2.5 Pro and Anthropic's Claude 3.7 Sonnet, both of which are nearly unmatched on coding benchmarks.
OpenAI claimed that GPT-4.1 outperforms its predecessors, the GPT-4o and GPT-4o mini, in several benchmarks, including SWE-bench, related to software engineering. However, it falls short of coming on par with Gemini 2.5 and Claude 3.7 in benchmark scores.
All of the three new GPT models come with varied speeds and costs, with GPT-4.1 nano being the fastest and most affordable and GPT-4.1 being the most capable. The prices of these coding-oriented AI models start from $0.10 and go up to $2 per million input tokens.
Even after registering notable upgrades in formatting and tool usage, GPT-4.1’s accuracy is said to be dropping when dealing with larger inputs.
