OpenAI has introduced an early beta of Canvas, a new interface for working with ChatGPT on writing and coding projects, including a separate window to overcome the limitations of a chat-based user interface.

Canvas is built with GPT-4o (GPT 4 Omni), a model made available in May as a single model for text, vision and audio. 

According to OpenAI, the original chat interface that gives ChatGPT its name is “limited when you want to work on projects that require editing and revisions,” a category that includes most programming tasks. ChatGPT with Canvas will behave more like a chat UI alongside an editor, we’re told, with features including code review with inline suggestions, detecting and fixing bugs, and porting code to JavaScript, TypeScript, Python, Java, C++ or PHP.

The Canvas window can be opened manually or may pop up automatically according to triggers developed by OpenAI. Recognizing that having Canvas appear could be distracting, the OpenAI team said that “we intentionally biased the model against triggering to avoid disrupting our power users,” though this could change based on feedback.

Despite being early in its development, OpenAI is already rolling out Canvas. Subscribers to ChatGPT Plus and Team can use it now; enterprise and educational users will get access this week, and free users when it is out of beta.

The market for coding assistants is evolving rapidly but ChatGPT remains the most popular, according to a StackOverflow survey conducted between May and June. ChatGPT was used by 81.7 percent of developers surveyed versus 44.2 percent for second-placed GitHub Copilot and 22.4 percent for Google Gemini – percentages sum to more than 100 because of developers using more than one tool. Factors in the widespread use of ChatGPT are that it was early to the market, and still has a free option. 

An early user said: “I just tried it … a huge step up in having the discussion and then a separate canvas where the work takes place, and then iterations are done to the canvas.”

There is a question though about whether OpenAI is losing focus by positioning itself as both an API provider for other tools, and as an end-user service for developers and other users. The browser-based user interface is also limiting, compared to integration with an editor such as Visual Studio Code (VS Code).