Artificial Analysis Coding Agent Index adds refusal timing and fallback model views
Artificial Analysis says the timing view separates refusals from the task prompt alone from those after work began; the fallback view includes blocked attempts.
TLDR
Artificial Analysis says its Coding Agent Index now shows when refusals occur and which models agents switch to afterward. As of October 1, Claude Code with Sonnet 5.5 (max) ranked first, with a 4.5% safety refusal rate versus 8.9% for Claude Code with Opus 5.5 (max). Around 94% of Sonnet 5.5's refusals occurred after the first turn; the agent almost always switched to Opus 4.8. Provider-configured behavior may change.
Combined views
20K
2 Sources, first seen 8h ago
