Electricity Bench

claude-code + claude-opus-4-8 vs grok-cli + grok-4.5

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.

claude-cliclaude-code + claude-opus-4-8overall pendinggrok-cligrok-cli + grok-4.5overall pending
structured-extraction@v1AvsA
Dimensionclaude-code + claude-opus-4-8grok-cli + grok-4.5Delta
Capability100%100%even
Median latency4752 ms4580 ms172 ms · grok-cli + grok-4.5
Reliability100%100%even
Error rate0%0%even

claude-code + claude-opus-4-8: 2.0% of Claude Max 5x per run, about 50.0 runs per 5h. grok-cli + grok-4.5: 0.3% of SuperGrok per run, about 317.4 runs per week. 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.