OpenAI releases GPT-4.1 with major gains and refreshed knowledge cutoff

ChatGPT developer OpenAI has introduced three new large language models (LLMs) for the developer community: GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano.

According to OpenAI, these three models outperform GPT-4o and GPT-4o mini across the board, especially in coding and instruction following.

Another significant gain compared to its predecessors is that the new LLMs have larger context windows that support up to one million tokens of context. This allows users to add more information when filling out a prompt.

For comparison: GPT-4o models could handle up to 128,000 tokens of context. One million tokens are more than eight copies of the entire React codebase. Such a long context is great for processing large codebases or long documents.

GPT-4.1 excels in numerous industry standard benchmarks. It scores 54.6% on SWE-bench Verified, which is an improvement of 21.4% over GPT-4o. OpenAI’s leading model also scores better compared to its predecessors on instruction following and long context.

OpenAI claims GPT-4.1 mini is a significant leap in small model performance, beating GPT-4o in many benchmarks while reducing latency nearly by half and reducing cost by 83%. GPT-4.1 nano is OpenAI’s fastest and cheapest model and is ideal for tasks like classification and autocompletion.

“While benchmarks provide valuable insights, we trained these models with a focus on real-world utility. Close collaboration and partnership with the developer community enabled us to optimize these models for the tasks that matter most to their applications,” OpenAI says in a press release.

For example, Thomson Reuters tested GPT-4.1 with their professional AI assistant for legal work called CoCounsel. Multi-document review accuracy improved by 17% and the press agency found the LLM to be highly reliable at maintaining context across sources and accurately identifying nuanced relationships between documents.

GPT-4.1, mini, and nano feature a refreshed knowledge cutoff of June 2024. OpenAI promises that the chatbots will unlock “new possibilities for building intelligent systems and sophisticated agentic applications.” All three LLMs are now available for developers.

In addition, OpenAI is still working on its next major LLM called GPT-5. Sam Altman, co-founder and CEO of OpenAI, recently confirmed that the chatbot will be launched “in a few months.”