Electricity Bench

antigravity-cli + gemini-3.1-pro vs cursor-cli + composer-2.5 (Cursor Pro)

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

antigravity-cliantigravity-cli + gemini-3.1-proSolves up to tier 5 at 16.8% of Google AI Pro per run · 5.9 runs/weekoverall pendingcursor-clicursor-cli + composer-2.5 (Cursor Pro)Solves up to tier 5 at 1.7% of Cursor Pro per run · 60.2 runs/monthoverall pending
real-world@v3DvsC
Dimensionantigravity-cli + gemini-3.1-procursor-cli + composer-2.5 (Cursor Pro)Delta
Capability47%57%10% · cursor-cli + composer-2.5 (Cursor Pro)
Median latency150525 ms117231 ms33294 ms · cursor-cli + composer-2.5 (Cursor Pro)
Reliability100%100%even
Error rate0%0%even

antigravity-cli + gemini-3.1-pro: 16.8% of Google AI Pro per run, about 5.9 runs per week. cursor-cli + composer-2.5 (Cursor Pro): 1.7% of Cursor Pro per run, about 60.2 runs per month. 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.