Electricity Bench

2026-W39: Real-world issues

Week 39 of 2026 graded 13 coding agents on real-world issues. GPT-5.6 Sol leads with C-. The biggest move since W38 2026 is Gemini 3.8 Flash, D+ to F. Grades compare only within the same suite version, and plan usage is each agent's share of its own subscription. Every agent ran the same suite cut headless through its own tools.Measured

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

Published suite snapshot for 13 graded agents, current through Sep 21, 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
GPT-5.6 Solcodex-cliC-60.0%tier 4 · partial tier 53m 13s100%2%/1wChatGPT Pro 5x
Claude Opus 5claude-codeC-56.7%tier 2 · partial tier 51m 56s100%21%/5hClaude Max 5x
Kimi K3 256Kkimi-cliC-56.7%tier 2 · partial tier 54m 5s100%115%/5hKimi Allegretto
Gemini 3.1 Proantigravity-cliD+53.3%tier 3 · partial tier 52m 55s100%9.8%/1wGoogle AI Pro
Claude Fable 5.1claude-codeD+53.3%tier 3 · partial tier 52m 20s100%100%/5hClaude Max 5x
Claude Haiku 4.5claude-codeD+50.0%tier 2 · partial tier 53m 39s100%13%/5hClaude Max 5x
Claude Sonnet 5claude-codeD+50.0%tier 2 · partial tier 52m 55s100%16%/5hClaude Max 5x
GPT-5.6 Terracodex-cliD+50.0%tier 2 · partial tier 53m 29s100%2%/1wChatGPT Pro 5x
Composer 2.5cursor-cliD+50.0%tier 2 · partial tier 51m 21s100%1.7%/1mCursor Pro
Grok 4.6grok-cliD46.7%tier 4 · partial tier 57m 10s100%3%/1wSuperGrok Plus
GPT-5.6 Lunacodex-cliD45.0%tier 1 · partial tier 53m 7s100%0%/1wChatGPT Pro 5x
GPT-6 Astracodex-cliD43.3%tier 2 · partial tier 52m 15s100%5%/1wChatGPT Pro 5x
Gemini 3.8 Flashantigravity-cliF28.3%none3m 11s100%18.0%/1wGoogle AI Pro

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.