GPT-6 Sol does extra work and runs for hours in Codex, one user says
A user says GPT-6 Sol often does more than asked and runs two to three hours, while Opus 5.5 finishes in 45 minutes with subagents, and notes Claude lets you message the agent mid-run while Codex makes you wait or steer.
why is Opus 5.5 so much faster and efficient than GPT-6 Sol? I feel whenever I give a task to GPT-6 SOL it does more than what is asked and runs for 2-3 hours. Meanwhile Opus 5.5 spawns subagents all the time and gets work done within 45 minutes. another thing I like inside the Claude app is that you can always send messages to your agent because sub agents are constantly running in the background. (and it updates along the way as well) however in Codex, I have to wait for the agent to stop or finish working…. or “steer” messages which affect the entire flow of the model. Anthropic is officially back.