Electricity Bench

gemini-3.1-pro vs kimi-k3-256k

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 runFOVERALLkimi-k3-256kClears tier 2 (partial tier 5) at 99.0% of Kimi Allegretto per runDOVERALL
real-world@v3DvsC
Dimensiongemini-3.1-prokimi-k3-256kAdvantage
Capability53%57%+3% (kimi-k3-256k leads)
Median task time3m 25s3m 3s22s faster (kimi-k3-256k leads)
Approx. task cost$0.026 / task$0.018 / task$0.009 / task lower (kimi-k3-256k leads)

gemini-3.1-pro: 8.5% of Google AI Pro (51.2% / 5h, 8.5% / 1w) per run. kimi-k3-256k: 99.0% of Kimi Allegretto (99.0% / 5h, 20.0% / 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-prokimi-k3-256kAdvantage
Capability16%51%+35% (kimi-k3-256k leads)
Median task time1m 47s5m 16s3m 28s faster (gemini-3.1-pro leads)
Approx. task cost$0.009 / task$0.034 / task$0.025 / 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. kimi-k3-256k: 51.0% of Kimi Allegretto (51.0% / 5h, 10.0% / 1w) per run. Each figure is a share of that agent's own plan, so there is no delta to draw.

vibe-coding@v1FvsD
Dimensiongemini-3.1-prokimi-k3-256kAdvantage
Capability23%38%+15% (kimi-k3-256k leads)
Median task time4m 39s6m 37s1m 59s faster (gemini-3.1-pro leads)
Approx. task cost$0.039 / task$0.027 / task$0.012 / task lower (kimi-k3-256k leads)

gemini-3.1-pro: 0.9% of Google AI Pro (5.1% / 5h, 0.9% / 1w) per good solve. kimi-k3-256k: 10.2% of Kimi Allegretto (10.2% / 5h, 2.0% / 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 kimi-k3-256k?

gemini-3.1-pro clears higher: tier 3 of 5 versus tier 2 for kimi-k3-256k on real-world@v3, our private suite of real-world coding problems. Latest measurement 2026-08-17.