OpenAI today announced the launch of three new GPT models that are available through the OpenAI API. Called GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, the models are not available for ChatGPT at the current time.
GPT-4.1 includes major improvements in coding, instruction following, and long context, according to OpenAI. The models outperform GPT-4o and GPT-4o mini in all tasks, especially coding tasks. The new models support up to one million tokens of context, and can use that for better long-context comprehension.
On SWE-bench Verified, GPT-4.1 showed a 21.4 percent improvement over GPT-4o for coding, and a 26.6 percent improvement over GPT-4.5. The slimmed down GPT-4.1 mini matches or exceeds GPT-4o with reduced latency and an 83 percent lower cost. GPT-4.1 nano is OpenAI's fastest and cheapest model, ideal for classification, autocompletion, and similar tasks.
OpenAI says that many of the improvements in GPT-4.1 have already been incorporated into the ChatGPT version of GPT-4o, with more functionality to be added in the future. All GPT-4.1 models have a cutoff date of June 2024, which means they are up to date on events that took place before then.
With the introduction of GPT-4.1, OpenAI is deprecating GPT-4.5 in the API, as GPT-4.1 offers much of the same functionality at a lower cost.
GPT-4.1 costs $2 per million input tokens and $8 per million output tokens, while GPT-4.1 mini costs $0.40 per million input tokens and $1.60 per million output tokens. GPT-4.1 nano is just $0.10 per million input tokens and $0.40 per million output tokens. Customized models are available at a higher price.
