Claude Opus 4.7 officially released! Benchmark crushes 4.6 + GPT-5.4, visual resolution improved by 3x+, more rigorous on long tasks, more precise instruction execution, and even self-verifies outputs... API now adds xhigh effort + task budget, and /ultrareview in Claude Code is just next level. Anthropic really threw the 'hardest jobs' at us this time. Is the era of less micromanagement finally here?!