OpenAI Unveils Coding-Focused GPT-4.1 While Phasing Out GPT-4.5
OpenAI has announced the introduction of its GPT-4.1 model family, which emphasises improved coding capabilities and a significant context window expansion to 1 million tokens. This model structure includes standard, mini, and nano variants available via API, with plans to phase out GPT-4.5 due to performance evaluations.
Key Points
- GPT-4.1 prioritises coding performance and has a context window of approximately 750,000 words.
- It includes three variants: standard, mini, and nano, with the nano being the most affordable at $0.10 per million input tokens.
- GPT-4.1 scored 54.6% on SWE-bench Verified, slightly lower than competitors Google and Anthropic.
- The model is set to replace GPT-4.5, which was released only two months prior, as GPT-4.1 delivers similar or better performance at lower costs.
- All versions feature a knowledge cutoff of June 2024, offering more up-to-date contextual understanding.
Why should I read this?
This article highlights a significant shift in OpenAI’s offerings, focusing on enhancing capabilities for coding professionals. With GPT-4.1, OpenAI aims to remain competitive in the rapidly evolving AI landscape, making it crucial for developers and tech enthusiasts to stay informed about tools that can impact software development and engineering practices.
“`