← sources

Is the current Opus 5.5 model back on track?

source: reddit (reddit:ClaudeAI)published 2026-09-30 06:10ingested 2026-10-01 02:38

retained text

Hi guys,

I had been using Claude Cowork to make progress through my work faster. I do biological research, and I mainly used it to help me source a bunch of literature, automate my analytical workflows, and find gaps in my own reasoning. During the Opus and Sonnet 4.6–4.8 models, they worked great. But with the later models, I found that they became obviously lazier (thinking session also stopped showing what they were trying to do). They stopped searching and kept reusing materials/information I had provided repeatedly, even when they were no longer appropriate for the task. It was kinda stuck in a loop, and starting new tasks didn't seem to work.

From a suggestion in this sub, I moved on to GPT Work with Luna and Sol. All the issues I previously had with the later Claude models were solved.

Now that my subscription to GPT has ended, and Opus 5.5 just came out, I'd like to hear your opinions about the performance of these two on non-coding tasks. Has Opus 5.5 regained whatever it lost previously?

Thank you!

document fetched 2026-10-01 00:25 · canonical https://reddit.com/r/ClaudeAI/comments/1wtxcin/is_the_current_opus_55_model_back_on_track