evals.report
BenchmarksLabsCompareRun guidesIn the wild
BenchmarksCoding

Vibe Code Bench

An end-to-end web application development benchmark (by Vals AI / Replit) where models build complete full-stack web apps from natural-language specifications in a sandboxed environment with production services (Supabase, Stripe, email), then are scored by an autonomous browser agent on overall application pass accuracy.

CodingOverall accuracyHigher is better
ScoresAboutRun this benchmark

What is Vibe Code Bench?

An end-to-end web application development benchmark (by Vals AI / Replit) where models build complete full-stack web apps from natural-language specifications in a sandboxed environment with production services (Supabase, Stripe, email), then are scored by an autonomous browser agent on overall application pass accuracy. evals.report tracks reported Vibe Code Bench scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.

Top reported Vibe Code Bench score: Claude Opus 4.8 — 82.72% (Overall accuracy).

ModelLabScore↓Source modelStatusDate
Claude Opus 4.8Anthropic82.72%—OfficialMay 28, 2026Details
Claude Opus 4.7Anthropic71.00%—OfficialApr 16, 2026Details
GPT-5.5OpenAI69.85%—OfficialApr 23, 2026Details
GPT-5.4OpenAI67.42%—OfficialMar 5, 2026Details
GPT-5.3-CodexOpenAI61.77%—OfficialFeb 5, 2026Details
Claude Opus 4.6Anthropic57.57%—OfficialFeb 5, 2026Details
GPT-5.2OpenAI53.50%—VerifiedDec 11, 2025Details
Claude Sonnet 4.6Anthropic51.48%—VerifiedFeb 17, 2026Details
DeepSeek V4 ProOpenDeepSeek49.93%—VerifiedApr 24, 2026Details
Gemini 3.5 FlashGoogle DeepMind48.68%—VerifiedMay 19, 2026Details
MiniMax M3OpenMiniMax47.57%—OfficialJun 1, 2026Details
GPT-5.2-CodexOpenAI37.91%—VerifiedDec 18, 2025Details
Kimi K2.6OpenMoonshot AI37.89%—VerifiedApr 20, 2026Details
Gemini 3.1 Pro PreviewGoogle DeepMind32.03%—VerifiedFeb 19, 2026Details
GLM-5.1OpenZ.ai31.46%—VerifiedApr 7, 2026Details
Qwen 3.6 PlusAlibaba / Qwen25.57%—VerifiedApr 2, 2026Details
GPT-5.1OpenAI24.61%—VerifiedNov 12, 2025Details
GLM-5OpenZ.ai23.36%—VerifiedFeb 11, 2026Details
Claude Sonnet 4.5Anthropic22.62%—VerifiedSep 29, 2025Details
Claude Opus 4.5Anthropic20.63%—VerifiedNov 24, 2025Details
Gemini 3 FlashGoogle DeepMind20.20%—VerifiedDec 17, 2025Details
GPT-5OpenAI20.09%—VerifiedAug 7, 2025Details
Muse SparkMeta19.67%—VerifiedApr 8, 2026Details
Grok 4.3xAI19.40%—VerifiedApr 17, 2026Details
Kimi K2.5OpenMoonshot AI17.54%—VerifiedJan 27, 2026Details
MiniMax M2.5OpenMiniMax14.85%—VerifiedFeb 12, 2026Details
Gemini 3 ProGoogle DeepMind14.30%—VerifiedNov 18, 2025Details
GPT-5 miniOpenAI14.17%—VerifiedAug 7, 2025Details
MiniMax M2.7OpenMiniMax11.93%—VerifiedMar 18, 2026Details
Qwen3 MaxAlibaba / Qwen3.51%—VerifiedSep 5, 2025Details
GLM-4.6OpenZ.ai3.09%—VerifiedSep 30, 2025Details
Grok 4.1 fast reasoningxAI1.20%—VerifiedNov 19, 2025Details
Gemini 2.5 ProGoogle DeepMind0.40%—VerifiedMar 25, 2025Details

Each row reports the model’s Overall accuracy on Vibe Code Bench. Click a row for the full run context.