tech

April 2, 2026

Vibe Check: Opus 4.6—The Best Coding Model We’ve Tested (With Some Maddening Habits)

It one-shotted a problem other models missed—and brings agentic, parallel work to non-coding tasks

Vibe Check: Opus 4.6—The Best Coding Model We’ve Tested (With Some Maddening Habits)

TL;DR

  • Opus 4.6 is the best AI coder tested, outperforming GPT-5.3 Codex and Opus 4.5 on a challenging iOS coding task.
  • The model is more thorough, explores context more carefully, and is smarter than Opus 4.5, but is slower and more verbose.
  • Opus 4.6 offers a more fluid drafting experience for writing and editing, applying editorial rules more consistently.
  • It is Anthropic's most agentic model yet, driving tasks forward with less handholding through parallelization and autonomous actions.
  • "Adaptive Thinking" replaces "Extended Thinking," adjusting reasoning based on task difficulty.
  • Users report successes in shipping code fixes for complex iOS problems and appreciate the default parallelization for knowledge-work tasks.
  • Despite improvements, Opus 4.6 can still make unexpected changes and exhibit 'AI-isms' in writing.
  • The model's strengths include early scope locking, parallel threads by default, and adaptation to user intent.
  • Weaknesses include wavering pace under load, uneven surface alignment in visual builds, and occasional signal haze.