Report
Inside AI coding in 2026
Last updated: 18th of June 2026
Last February, we released AI coding tracking as part of cubic's code review CLI. It allows users to track which lines of code were written by AI, including the model and coding agent.
We know most code is written by AI nowadays, but how much? Which models are most popular? More importantly, how do they perform?
As the #1 code reviewer, we parse thousands of commits a day by the best-performing AI engineering teams like Resend, n8n, Legora, and many more. We've analysed our data to shed some light on this topic.*
TL;DR
GPT-5.5 produces fewer bugs than any other model.
Despite this, Claude Code is the most popular coding agent, and the Opus family of models is used by 79% of the analysed teams.
Claude Fable was the fastest-adopted model in cubic's history, adopted by 30% of the analysed teams in the first 48 hours of its release.
GPT-5.5 produces fewer bugs than any other model.
Model quality: GPT-5.5 produces fewer bugs than any other model
Models are continuously getting better at coding, and that comes with a steady decrease in bugs. Each new model produces code with fewer bugs than the previous model (Opus 4.7 generated half the bugs of its previous version, Opus 4.6), except for Opus 4.8, which has a surprisingly high bug rate.
GPT-5.5 produces fewer bugs than flagship Anthropic models and is the least buggy model.
AI has tripled developer output
The typical developer has roughly doubled the number of merged PRs over the past year, from about 6 per month in mid-2025 to around 12 today. Those PRs are also 2.2× larger at the median.
In summary, developers are merging roughly three times more code than they were a year ago, a substantial increase in developer output. AI tends to generate more verbose and boilerplate-heavy code than humans, but although the impact is difficult to measure precisely, the value delivered has clearly increased with the adoption of AI.
Developers merged a median of 5820 lines of code changes in May 2026, compared to 1930 in June 2025, a year ago. That's a 3x increase.
Total numbers of PRs merged doubled in the past year, from 6 per month (June 2025) to 12 (May 2026).
PRs are getting bigger too, from a median of 56 lines of code changed per PR in June 2025 to 124 in May 2026, a 2.2x increase.
Coding agents: Claude dominates
Claude Code is the dominant coding agent, used by 80% of developers, with consistently high adoption. Cursor is used by 18% of developers, generally in combination with another agent, but is losing market share every week, while Codex maintains a stable 17% market share.
Model usage: new frontier models win
Anthropic's Opus model family is clearly the most popular among developers, with 80% of developers using it as their primary model every week. OpenAI's GPT family comes in second and has seen increased adoption following the release of GPT-5.5.
The week-over-week data reveals a clear pattern: new flagship frontier models are quickly adopted within two weeks of release.
The short-lived story of Fable
The purple blob in the model-share chart is Claude Fable, the latest model from Anthropic. It was released on Tuesday, June 9, and by Thursday it had been adopted by 30% of developers on the platform, the fastest adoption of any model in cubic's history.
On Friday of that week, the U.S. government banned it, and Anthropic took it down. Despite being available for only half the week, it captured 15% market share during that period.
Are humans writing code at all?
Among the code we track, AI authors the vast majority of it (90–100%). This has gone up from 78% just two months ago (mid-April), and the upward trend is clear.
You can track AI coding activity and attribution for your team using the cubic CLI.
* The dataset is based on hundreds of thousands of commits made by teams of at least 2 developers. These metrics focus exclusively on commits where AI attribution was tracked via our CLI, representing only a fraction of all commits. We started tracking this data around mid-February. While this doesn't represent the total codebase, it offers a compelling directional look at emerging patterns.