Alibaba's Qwen AI model trained to write Clojure code rivals GPT-4 performance in programming benchmarks

Alibaba trained its Qwen AI model (language model that understands and generates text) specifically on the Clojure programming language and claims it now matches or exceeds GPT-4 (OpenAI's previous flagship model) at writing Clojure code, a less common language that few AI models optimize for.

The difference: Qwen was fine-tuned (specialized training focused on one task) exclusively for Clojure, whereas general-purpose AI models like GPT-4 were trained on all programming languages equally. This shows that an AI model trained narrowly on one language can outperform broader competitors at that specific task.

For Clojure developers, this means a new alternative to GPT-4 and GPT-5 that understands the language's unique syntax better—code suggestions should require fewer rewrites and save time on boilerplate (repetitive code). For companies evaluating AI coding assistants, this demonstrates Alibaba can compete with OpenAI in specialized domains rather than just general tasks.