OpenAI has unveiled a new family of AI models aimed at coding and competing with the likes of Google and Anthropic, according to a Wired article.
OpenAI is releasing three models, GPT 4.1, GPT 4.1 Mini, and GPT 4.1 Nano, which it claims outperform its previous iterations in some aspects and will be available to developers via API.
GPT 4.1 scored 55% on SWE-Bench, a benchmark for coding, and OpenAI claimed it is 40% faster than its previous model and 80% cheaper for users.
Anthropic and Google have both launched models that can write code, and OpenAI’s latest offering includes the ability to analyse eight times as much code and follow instructions more effectively.
OpenAI wants to show it can achieve a profitable business amongst heavy competition and plunging profits, having announced last week it would offer an open-weight model developers can modify for free during the summer.
