OpenAI has launched its new AI model, GPT-4.1, which operates at a PhD level and features significant improvements in coding and context handling, including a larger context window of up to 1 million tokens. The model, along with its Mini and Nano versions, is already being used for scientific discoveries, and developers can earn tokens by providing feedback, while a new IDE called Windsurf offers free access to explore its capabilities.
OpenAI has introduced its latest AI model, GPT-4.1, which is rumored to cost around $20,000 per month and is designed to function at a PhD level. This model is part of a new family that includes GPT-4.1 Mini and GPT-4.1 Nano, with early previews already being utilized for scientific discoveries and innovative ideas. Additionally, OpenAI has released a prompting guide specifically for GPT-4.1, which can enhance performance by up to 20% through optimized prompting techniques.
The term “Quazar” was revealed during the announcement, referring to a powerful and distant active galactic nucleus powered by a supermassive black hole. It appears that Quazar may be linked to the GPT-4.1 Mini model, which is now accessible via the OpenAI API and the OpenAI Playground. The live stream event highlighted the capabilities of the new models, particularly in coding, where GPT-4.1 scored 54.6 on the SUI Bench Verified, marking a 21.4% improvement over its predecessor, GPT-4.0.
Developers using the APIs can earn free OpenAI tokens by providing feedback and sharing prompts with the company. This initiative allows users to gain up to 1 million tokens per day for the GPT-4.1 preview and even more for the Mini and Nano versions. Additionally, a new integrated development environment (IDE) called Windsurf will offer free unlimited access to GPT-4.1 for a limited time, encouraging users to explore its capabilities.
The performance of the new models shows significant improvements, particularly in instruction following and context handling. GPT-4.1 is noted for its ability to follow instructions more reliably and has a larger context window of up to 1 million tokens, which enhances its performance in tasks requiring extensive context. OpenAI has also developed a new benchmark for evaluating multi-round co-reference tasks, demonstrating the model’s ability to maintain context across multiple requests.
Overall, while the release of GPT-4.1 may not seem groundbreaking, it represents a step forward in AI capabilities, particularly in coding and scientific research. Early users have reported that these models could significantly enhance their ability to conduct experiments and discover new materials. The anticipation for future models, including the full 03 and 04 families, suggests that even more advanced capabilities are on the horizon, potentially revolutionizing various fields of research and development.