Do more expensive AI coding models write fewer bugs?

As of Sep 26, 2026, the most expensive models, Claude Fable 5, GPT-6 Astra, and Claude Fable 5.1 ($50 per 1M output tokens), had relative bug rates of 0.70, 1.16, and 1.17. The lowest rate belonged to GPT-5.5 (0.47, $30 per 1M output tokens).

Updated Sep 26, 2026

AI coding model bug rate vs price

As of Sep 26, 2026, GPT-5.5 had the lowest relative bug rate (0.47, $30 per 1M output tokens). The cheapest model with a below-average bug rate was GPT-5.6 Sol (0.73, $20 per 1M output tokens).

Lower is better

Rates and list prices

USD per 1M tokens

Ranked models by relative bug rate with input and output list prices per 1M tokens.
Model	Rate	Input	Output
GPT-5.5	0.47	$5	$30
Claude Fable 5	0.70	$10	$50
GPT-5.6 Sol	0.73	$4	$20
Claude Opus 4.8	0.88	$5	$25
GPT-6 Astra	1.16	$10	$50
Claude Fable 5.1	1.17	$10	$50
GPT-5.6 Luna	1.28	$0.20	$1.20
Claude Opus 5	1.58	$5	$25
Claude Sonnet 5	2.10	$2	$10

Questions and answers

Do more expensive AI coding models write fewer bugs?: As of Sep 26, 2026, the most expensive models, Claude Fable 5, GPT-6 Astra, and Claude Fable 5.1 ($50 per 1M output tokens), had relative bug rates of 0.70, 1.16, and 1.17. The lowest rate belonged to GPT-5.5 (0.47, $30 per 1M output tokens).