SemiAnalysis has analyzed 2.27 million Claude responses to track tool usage across different models. Their findings indicate a downward trend in tool calls from Claude Opus 4.6 to Opus 4.8. However, Claude Fable 5 deviates from this trend, exhibiting a higher average of 1.00 tool calls per response, compared to Opus 4.8's 0.76 and Opus 5's 0.79. This suggests that Fable 5's tool usage frequency is closer to earlier Opus models, prompting speculation that its performance gains might be linked to more frequent and accurate tool utilization. AI
IMPACT Analysis suggests Claude Fable 5 may achieve performance gains through more frequent and accurate tool use.
RANK_REASON Analysis of model behavior from a third-party researcher.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
