OpenAI introduces its new multimodal model GPT-4.1 with expanded context and lower costs

OpenAI has introduced GPT-4.1, the latest iteration in its series of multimodal models, succeeding last year's GPT-4o. This new model offers enhanced capabilities in coding and instruction-following, with a significantly expanded context window of up to 1 million tokens, a notable increase from GPT-4o's 128,000-token limit. GPT-4.1 improves its ability to detect relevant information and filter out distractions across the full token context.

Developers now have access to GPT-4.1 along with two scaled-down versions: GPT-4.1 Mini and GPT-4.1 Nano. The Mini version provides a cost-effective option, continuing the trend of previous "Mini" models, while the Nano version stands as OpenAI's most compact, rapid, and economical offering. Additionally, GPT-4.1 is 26 percent less expensive to operate than its predecessor, GPT-4o.

The company has previously announced the retirement of the GPT-4 model from ChatGPT by April 30th, with GPT-4o serving as its successor. The company also announced that GPT-4.5 preview in the API will also be deprecated on July 14th, as GPT-4.1 delivers comparable or superior performance with reduced costs and latency.