The best AI for coding in 2025 (including a new winner - and what ...
Have you ever wondered what "ZDNET Recommends" really means? At ZDNET, our recommendations are the result of extensive testing, research, and comparison shopping. We collect data from top sources, including vendor and retailer listings, along with independent review sites. By analyzing customer reviews, we aim to understand the real-life experiences of product users.
When you make a purchase through a retailer after clicking on a link from our site, we may earn affiliate commissions. This helps us maintain our work without impacting the content we cover or the price you pay. It's important to note that neither ZDNET nor our team members receive compensation for these independent reviews. Our editorial team is dedicated to providing you, our readers, with accurate information and expert advice to assist you in making informed decisions on technology products and services.
Exploring AI for Coding
My journey into the world of chatbots and AI-assisted programming began with the release of OpenAI's ChatGPT. This led me to test 14 large language models (LLMs) in real-world scenarios. Two years later, I've discovered that not all chatbots excel at coding. In this article, I will share the performance of each LLM based on my tests.
Out of the 13 LLMs tested, only five chatbots stood out. ChatGPT Plus and Perplexity Pro, priced at $20 per month each, were among the top recommendations. While their free versions are functional, the paid subscriptions offer enhanced features. Additionally, products from Google and Microsoft also made the list.
Top Recommendations
ChatGPT Plus, in combination with GPT-4o, passed all tests with flying colors. Its dedicated app for Mac users provides a seamless coding experience. On the other hand, Perplexity Pro impressed with its ability to run multiple LLMs, although it fell short in terms of user authentication.
Gemini Pro 2.5 from Google showcased significant improvements compared to its predecessor. However, access limitations may pose challenges for users. Microsoft Copilot, once a lackluster performer, excelled in the latest tests, even with its free version.
Interestingly, the free version of Claude 4 Sonnet outperformed its paid counterpart, Claude 4 Opus. This unexpected outcome highlights the unpredictability of AI implementations.
Final Thoughts
While AI-driven coding assistance has its limitations, these tools can streamline certain aspects of programming tasks. It's crucial to choose the right chatbot based on your specific requirements and budget. As technology continues to evolve, we can expect further advancements in AI capabilities for coding applications.
For a detailed analysis of my coding tests and the performance of each chatbot, refer to the following articles: How I test an AI chatbot's coding ability and How to use ChatGPT to write code.
