OpenAI has officially unveiled GPT-4.1, a large language model that is the successor to the multimodal GPT-4o algorithm released last year. According to the company, the new AI model has a larger context window and generally outperforms GPT-4o “in almost every way.” In addition, the algorithm’s ability to write code and follow instructions has been improved.
Image source: Levart_Photographer / unsplash.com
GPT-4.1 is already available to developers, along with two smaller versions of the neural network. These are the GPT-4.1 Mini and GPT-4.1 Nano algorithms, which OpenAI says is the “smallest, fastest, and cheapest” AI model. All three versions of GPT-4.1 are multimodal, meaning they can work with other data, such as images or videos, in addition to text. The context window size has increased to 1 million tokens, which is significantly larger than GPT-4o’s 128,000 tokens.
GPT-4.1 is said to be able to process contextual information well throughout the entire user interaction. “We also trained it to be much more robust than GPT-4o in recognizing relevant text and ignoring distracting elements across long and short context spans,” OpenAI said in a statement. GPT-4.1 is also 26% cheaper than GPT-4o, which is especially important following the debut of Chinese company DeepSeek’s super-efficient AI model.
Image source: OpenAI
The launch of GPT-4.1 comes as OpenAI prepares to retire its two-year-old AI model GPT-4. According to the official announcement, GPT-4o will become the “natural successor” to GPT-4 after April 30. OpenAI will also close API access to the GPT-4.5 preview on July 14, as “GPT-4.1 offers improved or similar performance on many key features at significantly lower cost and overhead.”