Electricity Bench

gemini-3.1-pro vs gemini-3.7-flash

Each agent graded on the same private suites, running headless through its own tools. Advantages are oriented so a win is a win, whichever way the metric runs.

gemini-3.1-proClears tier 3 (partial tier 5) at 8.5% of Google AI Pro per runFOVERALLgemini-3.7-flashClears tier 3 (partial tier 5) at 11.3% of Google AI Pro per runCOVERALL
real-world@v3DvsD
Dimensiongemini-3.1-progemini-3.7-flashAdvantage
Capability53%53%Even
Median task time3m 25s2m 50s35s faster (gemini-3.7-flash leads)
Approx. task cost$0.026 / task$0.035 / task$0.008 / task lower (gemini-3.1-pro leads)

gemini-3.1-pro: 8.5% of Google AI Pro (51.2% / 5h, 8.5% / 1w) per run. gemini-3.7-flash: 11.3% of Google AI Pro (67.8% / 5h, 11.3% / 1w) per run. Each figure is a share of that agent's own plan, so there is no delta to draw.

spec-planning@v1FvsD
Dimensiongemini-3.1-progemini-3.7-flashAdvantage
Capability16%40%+23% (gemini-3.7-flash leads)
Median task time1m 47s1m 59s12s faster (gemini-3.1-pro leads)
Approx. task cost$0.009 / task$0.014 / task$0.005 / task lower (gemini-3.1-pro leads)

gemini-3.1-pro: 0.8% of Google AI Pro (4.6% / 5h, 0.8% / 1w) per run. gemini-3.7-flash: 1.2% of Google AI Pro (7.3% / 5h, 1.2% / 1w) per run. Each figure is a share of that agent's own plan, so there is no delta to draw.

vibe-coding@v1FvsB
Dimensiongemini-3.1-progemini-3.7-flashAdvantage
Capability23%79%+57% (gemini-3.7-flash leads)
Median task time4m 39s4m 7s32s faster (gemini-3.7-flash leads)
Approx. task cost$0.039 / task$0.029 / task$0.011 / task lower (gemini-3.7-flash leads)

gemini-3.1-pro: 0.9% of Google AI Pro (5.1% / 5h, 0.9% / 1w) per good solve. gemini-3.7-flash: 0.6% of Google AI Pro (3.7% / 5h, 0.6% / 1w) per good solve. Each figure is a share of that agent's own plan, so there is no delta to draw.

Grades and deltas are comparable only within a suite version. See methodology.

Frequently asked questions

Which is better for coding, gemini-3.1-pro or gemini-3.7-flash?

They are even on our ladder: both clear tier 3 of 5 at 53% capability on real-world@v3, our private suite of real-world coding problems. Latest measurement 2026-08-17.