OpenAI has launched its latest AI model series, GPT-4.1, introducing major improvements in coding, instruction following, and long-context comprehension. The new models—GPT-4.1, GPT-4.1 Mini, and GPT-4.1 Nano—are designed to outperform previous iterations, offering developers more efficient and cost-effective AI solutions. These models mark a significant step forward in OpenAI's ongoing effort to push the boundaries of AI capabilities, particularly in practical, real-world applications.

The standout feature of GPT-4.1 is its expanded context window, supporting up to 1 million tokens, significantly surpassing the 128,000-token limit of its predecessor, GPT-4o. This enhancement allows the model to process and understand larger datasets, making it particularly effective for complex tasks such as analysing extensive codebases or lengthy documents. The increased context window allows for more nuanced understanding and more accurate responses in complex scenarios.

GPT-4.1 delivers notable gains in coding performance, showing a 21% improvement over GPT-4o and a 27% increase compared to GPT-4.5. OpenAI CEO Sam Altman emphasised that the model was optimised for real-world utility, stating, "These models are great at coding, instruction following, and long context. Benchmarks are strong, but we focused on real-world applications, and developers seem very happy." This focus on practical application is a key differentiator for the GPT-4.1 series.

OpenAI has introduced three versions of GPT-4.1 to cater to different needs:

• GPT-4.1: The standard model with full capabilities.
• GPT-4.1 Mini: A cost-efficient version with reduced latency.
• GPT-4.1 Nano: The smallest, fastest, and most affordable model, ideal for classification and autocompletion tasks.

This tiered approach allows developers to choose the model that best fits their specific requirements and budget constraints, making advanced AI more accessible to a wider range of users.

As the AI arms race heats up, OpenAI has unveiled GPT-4.1—its latest move against rivals like Google and Anthropic in the battle to build the most advanced programming models. Google’s Gemini 2.5 Pro, which also features a massive 1-million-token context window, is already scoring high across key coding benchmarks. Meanwhile, Anthropic’s Claude 3.7 Sonnet and DeepSeek’s V3 model out of China are steadily gaining ground. OpenAI says GPT-4.1 brings serious upgrades over GPT-4o, especially when it comes to coding. It’s better at solving complex tasks agentically, handling frontend work, sticking to specific formats, making fewer unnecessary edits, and using tools with more consistency.

Alongside the launch, OpenAI is also phasing out GPT-4.5 Preview from its API—clearing the decks for what it clearly sees as the next generation of its coding-first AI. With GPT-4.1 now matching—or outperforming—GPT-4.5 across core benchmarks, and doing it faster and cheaper, OpenAI has announced that GPT-4.5 Preview will officially be retired on July 14, 2025. The timeline gives developers a fair runway to transition. GPT-4.5, initially launched as a research preview to test the limits of compute-heavy models, served its purpose by offering valuable insights through hands-on developer feedback.

OpenAI says the creativity, writing flair, humour, and nuance that users loved in 4.5 won’t be lost—they’re bringing those strengths into the next generation of API offerings. The GPT-4.1 family—featuring the base model, mini, and nano—is now live and available to all developers. Thanks to major efficiency gains in its inference systems, OpenAI has slashed prices across the board. GPT-4.1 is now 26% cheaper than GPT-4o for median queries, while GPT-4.1 nano claims the title of OpenAI’s "cheapest and fastest model ever."

Here’s the pricing breakdown:

• GPT-4.1 comes in at $2 per million input tokens and $8 per million output tokens.
• GPT-4.1 mini drops that to $0.40 and $1.60.
• GPT-4.1 nano takes it even further to just $0.10 and $0.40 respectively.

There’s more good news for devs. OpenAI is bumping its prompt caching discount to 75% (up from 50%) for repeated context queries, and long context requests won’t cost extra—just the standard per-token rate.

With the launch of GPT-4.1, OpenAI is phasing out older models, including GPT-4, which will be retired from ChatGPT by April 30, and GPT-4.5, which will be deprecated by July 14. This transition reflects OpenAI's commitment to providing the most advanced and efficient AI solutions to its users.