# Auditing a 200K-line codebase

An Anthropic-reported codebase audit illustrates a long, multi-step coding workload for Opus 5.5.

Canonical URL: https://aiidelist.com/blog/opus-5-5-codebase-audit

Language: en

Published: 2026-09-25

Updated: 2026-09-25

An Anthropic-reported codebase audit illustrates a long, multi-step coding workload for Opus 5.5.

## What the source shows

Anthropic describes a tester’s audit and fixes across roughly 200,000 lines of code, with a reported completion time below three hours. The report compares it with a much longer Opus 5 attempt.

## What to inspect

The headline time is not enough to reproduce the comparison. A useful evaluation needs the same repository revision, issue scope, test commands and definition of completion. Human review time and changes rejected after the session also matter.

## A useful way to try this

Use a bounded module in your own repository. Ask for a documented issue list, evidence for each finding and a small reviewed patch. Compare both models on the same starting revision, and count accepted fixes rather than files touched.

## Evidence and limitations

This is a vendor-published report, not an AI IDE List benchmark. The underlying environment and complete execution transcript are not available in this listing. The evidence panel above records the source type, settings and prompt attribution. AI IDE List has not independently reproduced this result.

[Open the source](https://www.anthropic.com/claude-opus-5-5).

[Explore Claude Opus 5.5 examples](/models/claude-opus-5-5/examples) or [compare model specifications and costs](/models/compare/claude-opus-5-5-vs-gpt-6-astra).