An enhanced coding capability is now available in GPT-4.1 from OpenAI

OpenAI has officially launched its latest generation of models with GPT-4.1, accompanied by lighter variants GPT-4.1 mini and nano, setting a new benchmark for coding-focused AI. Designed with developers in mind, the GPT-4.1 series supports a massive 1 million-token context and comes with improved long-context reasoning and an updated knowledge base that extends to June 2024. These models aim to handle intricate software development tasks that typically require seasoned engineers.

With the introduction of GPT-4.1, OpenAI takes a competitive leap against other industry players like Google and Anthropic. Google’s Gemini 2.5 Pro, which also supports a million-token context window, is making waves in coding benchmarks, while Claude 3.7 Sonnet from Anthropic and China’s DeepSeek V3 are also gaining recognition. However, GPT-4.1 is built to stand out-OpenAI has emphasized its enhanced capabilities in areas like format adherence, frontend generation, and reduced unnecessary edits. It also improves tool use and autonomous problem-solving, key for advanced programming tasks.

As part of this rollout, OpenAI is retiring the GPT-4.5 Preview model by July 14, 2025. GPT-4.5 served as a research-driven, transitional model that helped developers push the limits of large-scale AI. While it brought creativity, humor, and nuanced writing to the table, OpenAI affirms those qualities have been carried forward into GPT-4.1 and its variants, which now match or exceed 4.5’s performance at lower cost and faster speed.

The new models are already available to developers, and pricing reflects significant reductions due to gains in inference efficiency. The cost of GPT-4.1 is $2 for every million input tokens and $8 for each million output tokens. The mini version drops those costs to $0.40 and $1.60, while nano offers the most affordable rates at $0.10 for input and $0.40 for output tokens-earning the title of OpenAI’s fastest and cheapest model to date.

In addition, developers benefit from a 75% prompt caching discount for repeated queries, up from the previous 50%, and long-context requests are billed at regular token rates with no additional fees.