Electricity Bench

Gemini 3.8 Flash vs Claude Haiku 4.5

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 3.8 FlashClears tier 3 (partial tier 5) at 20.4% of Google AI Pro per runDOVERALLClaude Haiku 4.5Clears tier 1 (partial tier 5) at 4.0% of Claude Max 20x per runDOVERALL
real-world@v3DvsC
DimensionGemini 3.8 FlashClaude Haiku 4.5Advantage
Capability53%62%+8% (Claude Haiku 4.5 leads)
Median task time4m 14s3m 31s43s faster (Claude Haiku 4.5 leads)
Approx. task cost$0.063 / task$0.004 / task$0.059 / task lower (Claude Haiku 4.5 leads)

Gemini 3.8 Flash: 20.4% of Google AI Pro (122.4% / 5h, 20.4% / 1w) per run. Its weekly limit holds 6 of the week's 33.6 5h windows. Claude Haiku 4.5: 4.0% of Claude Max 20x (4.0% / 5h, 1.0% / 1w) per run. Its weekly limit holds about 7.2 of the week's 33.6 5h windows. Each figure is a share of that agent's own plan, so there is no delta to draw.

spec-planning@v1DvsF
DimensionGemini 3.8 FlashClaude Haiku 4.5Advantage
Capability39%25%+14% (Gemini 3.8 Flash leads)
Median task time2m 39s2m 9s30s faster (Claude Haiku 4.5 leads)
Approx. task cost$0.022 / tasknot measured-

Gemini 3.8 Flash: 1.9% of Google AI Pro (11.7% / 5h, 1.9% / 1w) per run. Its weekly limit holds 6 of the week's 33.6 5h windows. Claude Haiku 4.5: 0.0% of Claude Max 20x per run. Its weekly limit holds about 7.2 of the week's 33.6 5h windows. Each figure is a share of that agent's own plan, so there is no delta to draw.

vibe-coding@v1DvsF
DimensionGemini 3.8 FlashClaude Haiku 4.5Advantage
Capability52%21%+31% (Gemini 3.8 Flash leads)
Median task time7m 7s4m 18s2m 49s faster (Claude Haiku 4.5 leads)
Approx. task cost$0.046 / task$0.003 / task$0.043 / task lower (Claude Haiku 4.5 leads)

Gemini 3.8 Flash: 1.0% of Google AI Pro (6.0% / 5h, 1.0% / 1w) per good solve. Its weekly limit holds 6 of the week's 33.6 5h windows. Claude Haiku 4.5: 0.2% of Claude Max 20x (0.2% / 5h) per good solve. Its weekly limit holds about 7.2 of the week's 33.6 5h windows. 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 3.8 Flash or Claude Haiku 4.5?

Gemini 3.8 Flash clears higher: tier 3 of 5 versus tier 1 for Claude Haiku 4.5 on real-world@v3, our private suite of real-world coding problems. Latest measurement 2026-09-02.