o3 is excellent for precise, drop-in replacement fixes.

Clause Opus is great at expanding the "what-if" scenarios for a project but tends to generate bloated or unmanageable codebases.

Sonnet is strong in general problem-solving and structure.

Gemini has been hit or miss — at times, its performance matches o3, but it often seems throttled, suddenly degrading to behave more like 4o, no actual proof of this.