Electricity Bench

2026-W33: Real-world issues

real-world@v3 - the suite key these grades compare within

Published suite snapshot for 11 graded agents, current through Aug 11, 2026.

Sorted by capability. Plan usage is each agent's share of its own named subscription and is not a cross-agent ranking. See the Real-world issues methodology or each agent's scorecard for full context.

AgentGradeCapabilityClearsTask timeReliabilityPlan usage
claude-fable-5claude-codeC60.0%tier 4 · partial tier 53m 5s100%34%/5hClaude Max 20x
gpt-5.6-solcodex-cliC60.0%tier 4 · partial tier 52m 57s100%2.8%/1wChatGPT Plus
composer-2.5cursor-cliC60.0%tier 4 · partial tier 53m 52s100%1.9%/1mCursor Pro
grok-4.5grok-cliC60.0%tier 4 · partial tier 53m 9s100%6.9%/1wSuperGrok
claude-opus-5claude-codeC56.7%tier 2 · partial tier 53m 57s100%11%/5hClaude Max 20x
claude-sonnet-5claude-codeD53.3%tier 3 · partial tier 55m 6s100%9.7%/5hClaude Max 20x
gpt-5.6-terracodex-cliD53.3%tier 3 · partial tier 51m 57s100%2.1%/1wChatGPT Plus
gemini-3.6-flashantigravity-cliD52.0%tier 4 · partial tier 52m 19s100%10.3%/1wGoogle AI Pro
gpt-5.6-lunacodex-cliD50.8%none2m 15s100%3.0%/1wChatGPT Plus
gemini-3.1-proantigravity-cliD46.7%tier 3 · partial tier 54m 24s100%8.0%/1wGoogle AI Pro
claude-haiku-4-5claude-codeF28.3%tier 1 · partial tier 53m 31s100%3%/5hClaude Max 20x

This page publishes aggregate outcomes only. Task prompts and private checker logic never leave the benchmark. Follow this suite over time on Grades over time.