OpenAI unveils GPT-4.1 for developers

OpenAI has officially announced GPT-4.1, the next-generation language model set to replace last year’s multimodal GPT-4o.

According to the company, GPT-4.1 outperforms its predecessor “across nearly every metric,” including context handling, code generation, and instruction accuracy.

GPT-4.1 is now available to developers and comes in several versions: GPT-4.1, GPT-4.1 Mini, and GPT-4.1 Nano. The latter is described as the smallest, fastest, and most cost-efficient model in the lineup. All versions support multimodal capabilities, enabling them to process text, images, video, and other data formats.

One of the most significant upgrades is the expanded context window — now up to 1 million tokens, a dramatic increase from GPT-4o’s 128,000. This allows the model to retain and reason over much larger amounts of information in a single prompt, enabling more accurate handling of long-form or complex interactions.

GPT-4.1 is also 26% cheaper than GPT-4o, a key move amid intensifying competition from Chinese developers — particularly DeepSeek’s high-efficiency model. OpenAI says the new version is better at filtering relevant data and ignoring noise in dense prompts.

The launch comes as OpenAI begins phasing out GPT-4. As of April 30, GPT-4o will become the official successor, and access to GPT-4.5 via API will be discontinued on July 14. The company says GPT-4.1 offers comparable or better performance at a lower cost.