@teortaxesTex posted that TB 3.0 appears to reward both high intelligence and strong agentic behavior, adding that Opus 5 does not seem clearly better than Sol. @xlr8harder replied that Opus 5 can surface useful insights on problems where Sol stalls, yet it proves harder to steer and less consistent in follow-through. The reply noted that switching between the two models depending on the task might change which one appears stronger overall. Both comments treat direct head-to-head ranking as difficult to settle from available runs.