Gemini 3.1 Pro

Gemini 3.1 Pro is hard to recommend as a review seat: a measured review cost $0.2499, and it ranks 2 of 17 with 67% [54%–80%] recall.

On recall, Claude Opus 5.5 measurably beats it, and the run can't separate it from GPT-6.1 Sol or Gemini 4 Argon. Those two measurably beat 7 and 6 ranked models, this one 4. A Gemini 3.1 Pro review costs 6.2× a Sol review and 4.7× an Argon review. It also costs 88× a GPT-6 Luna review, with recall the benchmark cannot separate from this model's.

The list rate ($2.00 / $12.00 per 1M tokens) is part of the bill, but the larger part is how much it writes. The review-call price in the spec sheet assumes a typical reviewer's output, and it wrote several times that, and a review took 349s. One of the three judges is from Google, so its vote was dropped when scoring this model.

The upside is quiet output: 0.8 [0.4–1.2] false alarms per clean PR, fewer on average than Claude Sonnet 5.5 (2.0 [1.5–2.4], intervals do not overlap), and 89% of catches on the defect's own lines. That helps if noise is your main cost (see reducing false positives). Compare it with Gemini 4 Argon before choosing.