Which AI model writes the most bugs in Shell?

As of Nov 14, 2026, Claude Sonnet 5.5 had the highest relative bug rate in Shell code (2.05, 105% more counted bugs than the Shell average), statistically tied with GPT-6.1 Sol.

Updated Nov 14, 2026

Best AI model for Shell code

As of Nov 14, 2026, GPT-6 Astra had the lowest relative bug rate in Shell code (0.63, 37% fewer counted bugs than the Shell average), statistically tied with Claude Fable 5.1, Claude Opus 5.5, and GPT-6.1 Sol, among 5 ranked models.

Models ranked by relative bug rate in Shell code. Lower values are better.
Rank	Model	Rate
1= (statistically tied)	GPT-6 Astra	0.63
2= (statistically tied)	Claude Fable 5.1	0.68
3= (statistically tied)	Claude Opus 5.5	0.99
4= (statistically tied)	GPT-6.1 Sol	1.55
5= (statistically tied)	Claude Sonnet 5.5	2.05

1.00 = Shell average · lower is better · = statistically tied

Other languages

TypeScript
Python
JavaScript
SQL
Go
Rust
Swift
Ruby
Vue
Terraform
C#
PHP
Svelte
Java
Kotlin
Dart
Elixir

Questions and answers

Which AI model writes the most bugs in Shell?: As of Nov 14, 2026, Claude Sonnet 5.5 had the highest relative bug rate in Shell code (2.05, 105% more counted bugs than the Shell average), statistically tied with GPT-6.1 Sol.