OpenAI unveils GPT-4.1 series, promising better performance at lower cost

Key Takeaways

OpenAI unveiled the GPT-4.1 series promising major improvements in coding, instruction following, and context understanding.
The GPT-4.1 models offer reduced latency and improved cost efficiency compared to previous versions.

Share this article

OpenAI has officially released its next-generation model lineup—GPT-4.1, along with GPT-4.1 mini and GPT-4.1 nano—featuring major improvements in coding, instruction following, and context understanding.

The rollout follows a statement earlier this month from OpenAI CEO Sam Altman revealing the company’s plans to release the o3 and o4-mini models “in a couple of weeks.”

Shortly after his remarks, The Verge reported that OpenAI was preparing to launch its new AI models imminently. The effort is part of the company’s broader strategy to enhance its current offerings ahead of the anticipated GPT-5 release in 2025.

According to OpenAI, the new GPT-4.1 models outperform previous versions across key benchmarks. The flagship model, GPT-4.1, scores 54.6% on SWE-bench Verified for coding tasks—a 21.4 percentage point improvement over GPT-4o.

On instruction following, GPT-4.1 also shows major gains. It outperforms GPT-4o by 10.5 points on the MultiChallenge benchmark, and demonstrates better comprehension of complex, multi-step instructions, even when prompts include formatting, tone, or sequencing constraints.

When it comes to context handling, all three models can handle up to 1 million tokens, up from 128,000 in the previous generation. OpenAI says the models perform reliably across the full context window, retrieving and processing the relevant information with greater consistency.

“We trained these models with a focus on real-world utility,” the company stated in its announcement. “Close collaboration and partnership with the developer community enabled us to optimize these models for the tasks that matter most to their applications.”

Along with performance gains, the new models come with reduced pricing. GPT‑4.1 is 26% less expensive than GPT‑4o for median queries, and GPT‑4.1 nano is OpenAI’s fastest and most affordable model to date, according to the company.

The new models are now available to developers via OpenAI’s API. With GPT‑4.1 delivering better performance at lower cost, OpenAI announced it will stop supporting the GPT‑4.5 Preview on July 14.

Share this article