OpenAI Unveils New GPT-4.1 Models Designed for Coding
OpenAI Releases New GPT-4.1 Models
OpenAI has recently introduced three new models under the GPT-4.1 framework, which are accessible through the OpenAI API. These models are named GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano. Notably, they are currently not integrated with the ChatGPT platform.
Major Improvements in GPT-4.1 Models
The GPT-4.1 models come with significant advancements, particularly in areas such as coding capabilities, instruction following, and understanding long contexts. The new models have shown remarkable performance, surpassing their predecessors (GPT-4o and GPT-4o mini) across all tasks, especially in coding-related challenges.
Contextual Understanding
One of the standout features of the GPT-4.1 models is their capacity to handle up to one million tokens of context. This expanded capability allows for improved comprehension in scenarios that involve lengthy context. As a result, users can expect more accurate and relevant outputs from these enhanced models.
Performance Metrics
On the SWE-bench Verified testing platform, the GPT-4.1 model exhibited a substantial improvement of 21.4% in coding tasks compared to GPT-4o, and a 26.6% improvement over GPT-4.5. The streamlined version, GPT-4.1 mini, matches or outperforms GPT-4o while offering lower latency and a cost reduction of 83%. Meanwhile, GPT-4.1 nano emerges as OpenAI’s most affordable and quickest model, suitable for tasks involving classification, autocompletion, and similar applications.
Implications for GPT-4.5
With the advent of GPT-4.1, OpenAI has decided to phase out GPT-4.5 from its API offerings. The new GPT-4.1 models deliver comparable functionalities at a more economical price point, making them a more appealing choice for developers and businesses.
Pricing Structure
The pricing for the GPT-4.1 models is structured as follows:
- GPT-4.1: $2 per million input tokens and $8 per million output tokens.
- GPT-4.1 mini: $0.40 per million input tokens and $1.60 per million output tokens.
- GPT-4.1 nano: $0.10 per million input tokens and $0.40 per million output tokens.
Additionally, OpenAI offers customized models at a higher price, accommodating specific user needs.
Update on Future Functionalities
OpenAI indicates that many features from the GPT-4.1 improvements have already begun to appear in the ChatGPT version of GPT-4o, and more enhancements are expected to be rolled out in the future. Furthermore, all GPT-4.1 models are equipped with a cutoff date of June 2024, ensuring that their data is relatively current and includes knowledge of events that occurred before this date.
This launch not only enhances OpenAI’s offerings but also illustrates the company’s ongoing commitment to advancing artificial intelligence capabilities in a user-friendly and cost-effective manner. As these models gain traction, they are likely to become integral tools for developers and businesses alike.
