Reviewers split hard on Opus 5

Reactions to Opus 5 split fast, with four prominent voices in AI landing on different reads of the same model, per Techmeme's roundup.

For builders choosing a model, the split points to two different things to optimize for: peak capability, or reliability across a long task.

Musk grouped Opus 5 with Grok 4.5, calling the pair alone on the Pareto frontier. Aaron Levie called it a huge jump over Opus 4.8.

Ethan Mollick pushed back with something more specific. He found Opus 5 quirky: strong on shorter tasks, less ambitious once the work stretches out.

Bindu Reddy split the difference, ranking Opus 5 behind Fable in sheer brilliance but still a strict improvement over 4.8.

Mollick's read is the one worth acting on. If Opus 5's weak spot really is long-horizon ambition, the fix is in how you hand it work, not in waiting for a better model.

Harnesses that break a big goal into short, verified steps should get more out of Opus 5. Ones that hand it one open-ended task probably won't.

Give Opus 5 a genuinely open-ended, multi-day task and see if the quirky label holds. Nobody's run that test yet.