OpenAI Unveils GPT-4.1 Models With Enhanced Coding Skills And Contextual Understanding

OpenAI has launched three innovative artificial intelligence models, named GPT‑4.1, GPT‑4.1 mini, and GPT‑4.1 nano, which surpass their predecessors, GPT‑4o and GPT‑4o mini, in various aspects. These new models demonstrate remarkable advancements, especially in coding and following instructions.

They are designed to be accessible to developers through an application programming interface (API), marking a significant step forward in AI technology. However, it's important to note that these models won't be available through ChatGPT, emphasizing their specialized developer-focused utility.

Enhanced Performance and Contextual Understanding

The latest models boast enhanced capabilities, including a more extensive context window capable of supporting up to 1 million tokens of context. This improvement enables them to better comprehend and utilize long contexts. OpenAI highlighted, "They also have larger context windows—supporting up to 1 million tokens of context—and are able to better use that context with improved long-context comprehension. They feature a refreshed knowledge cutoff of June 2024." This feature is expected to significantly enhance their performance across various tasks.

In the realm of coding, GPT‑4.1 stands out for its proficiency. Compared to its predecessor, it shows a marked improvement in completing coding tasks, from agentically solving problems to ensuring consistent tool usage without making unnecessary edits. Its capabilities extend to frontend coding and reliably following diff formats, showcasing its versatility and improved precision in software development tasks.

Real-World Software Engineering Skills

When evaluated on SWE-bench Verified, which assesses real-world software engineering skills, GPT‑4.1 achieved a success rate of 54.6% in completing tasks. This is a significant leap from the 33.2% success rate of GPT‑4o. These figures underscore the model's enhanced ability to navigate code repositories, finalize tasks, and generate functional, test-passing code, pointing to its potential to revolutionize coding practices.

Developers now have access to GPT‑4.1, GPT‑4.1 mini, and GPT‑4.1 nano, thanks to OpenAI's efforts to make these models widely available. This accessibility is poised to catalyze innovation and efficiency in software development and other fields reliant on advanced AI capabilities.

Improved Affordability and Speed

OpenAI has also introduced efficiency improvements in their inference systems, leading to a reduction in prices for the GPT‑4.1 series. Notably, GPT‑4.1 is now 26% cheaper than GPT‑4o for median queries. Furthermore, the GPT‑4.1 nano has been unveiled as the cheapest and fastest model to date, promising to make cutting-edge AI technology more accessible to a broader range of developers.

The cost structure for these models has been clearly defined, with GPT-4.1 priced at $2 per million input tokens and $8 per million output tokens. The GPT-4.1 Mini offers a more affordable rate at $0.40 per million input tokens and $1.60 per million output tokens. The GPT-4.1 nano model, recognized as the most economical option, is priced at merely $0.10 per million input tokens and $0.40 per million output tokens. These pricing strategies are designed to encourage wider adoption and experimentation with these advanced AI models among the developer community.

In conclusion, OpenAI's introduction of the GPT‑4.1 series represents a significant milestone in the field of artificial intelligence. With enhanced performance, improved contextual understanding, and more accessible pricing, these models hold the promise of driving innovation across various sectors. As these tools become integrated into developers' workflows, we can anticipate a new era of AI-driven solutions and advancements.

Via