GPT 6.1 Sol and Claude Sonnet 5.5 split on speed, cost, and verification in agent tasks
Overview
Leiphone tested GPT 6.1 Sol and Claude Sonnet 5.5 through Claude Code on three agent tasks: a 3D escape room, a policy investigation, and a chemistry explainer.
Both models have the same per-token price on OpenRouter, but Sol's total cost was $38.5 against $10.9 for Sonnet 5.5, about 3.5 times higher. Leiphone attributes the gap to how the agents act, such as Sol archiving and cross-checking more sources, rather than to token prices.
Written by AI from the articles below · updated Nov 3, 12:40 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- LeiphoneNewsGPT 6.1 Sol and Claude Sonnet 5.5 split on speed, cost, and verification in agent tasks
AILeiphone tested GPT 6.1 Sol and Claude Sonnet 5.5 through Claude Code on three tasks: a 3D escape room, a policy investigation, and a chemistry explainer. Both models cost the same per token on OpenRouter, yet Sol's total bill was $38.5 against $10.9 for Sonnet 5.5, about 3.5 times higher. The author attributes the gap to differences in agent action paths, such as Sol's extra source archiving and cross-checking, rather than to the price per token.
Heat Trend
Not enough continuous observations to show a trend yet.