# Cookie-Bench: Continuous On-screen Key Interaction Evaluation for Web Generation (Cookie-Bench)

> A 1,000-query web development benchmark that judges generated front-end apps by deploying them and letting an evaluator agent use them, scoring functionality and aesthetics without reference implementations or test suites.

Canonical page: https://benchlm.ai/benchmarks/cookie-bench

- Category: [Coding](/coding)
- Last updated: November 13, 2026

## About Cookie-Bench

- Year: 2026
- Tasks: 1,000 web development queries (React setting)
- Format: Benchmark score (0-100)
- Difficulty: Interactive web apps and static pages across three tiers
- Paper: [Cookie-Bench: Continuous On-screen Key Interaction Evaluation for Web Generation](https://arxiv.org/abs/2605.30000)

Baidu researchers released Cookie-Bench in May 2026 with Cookie, its evaluator: 1,000 queries across 11 domains and 54 leaf categories, covering static pages and interactive applications, balanced across three difficulty tiers and three target-language groups, with briefs rewritten to resist recall. Cookie forms a first impression from the rendered page, then drives the app with a computer-using agent while recording screenshots, screen video and audio, and only then scores functionality and aesthetics. Baidu reports ERNIE 5.5 Flash Preview at 88.2 on the React setting in its November 2026 launch, next to its own runs of Claude Opus 5.5 (91.5), GPT-6.1 Sol (89.7), Hy4 (88.8), DeepSeek V4.1 Flash 1017 (87.9), GLM-5.3 (80.4) and GLM-5.3-Flash (76.9), all judged by Qwen3.8-Max. BenchLM stores these Baidu-run rows as display-only.

Cookie-Bench is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (7 models)

None of the 7 results on this page has a documented run by the benchmark or an independent evaluator: 1 reported by the lab that built the model and 6 reported in another publisher’s document.

| Rank | Model | Creator | Score | Who ran it |
|------|-------|---------|-------|-----------|
| 1 | [Claude Opus 5.5](/models/claude-opus-5-5) | Anthropic | 91.5% | Lab-reported, from another publisher’s document |
| 2 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | OpenAI | 89.7% | Lab-reported, from another publisher’s document |
| 3 | [Hy4](/models/hy4) | Tencent | 88.8% | Lab-reported, from another publisher’s document |
| 4 | [ERNIE 5.5 Flash Preview](/models/ernie-5-5-flash-preview) | Baidu | 88.2% | Reported by Baidu |
| 5 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | DeepSeek | 87.9% | Lab-reported, from another publisher’s document |
| 6 | [GLM-5.3](/models/glm-5-3) | Z.AI | 80.4% | Lab-reported, from another publisher’s document |
| 7 | [GLM-5.3-Flash](/models/glm-5-3-flash) | Z.AI | 76.9% | Lab-reported, from another publisher’s document |

## FAQ

### What does Cookie-Bench measure?

A 1,000-query web development benchmark that judges generated front-end apps by deploying them and letting an evaluator agent use them, scoring functionality and aesthetics without reference implementations or test suites.

### Which model scores highest on Cookie-Bench?

Claude Opus 5.5 by Anthropic currently leads with a score of 91.5% on Cookie-Bench.

### How many models are evaluated on Cookie-Bench?

7 AI models have published results on Cookie-Bench in the BenchLM catalog.

### Does Cookie-Bench affect BenchLM's overall score?

Not directly. Cookie-Bench is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Compare Top Models on Cookie-Bench

- [Claude Opus 5.5 vs GPT-6.1 Sol](/compare/claude-opus-5-5-vs-gpt-6-1-sol)
- [GPT-6.1 Sol vs Hy4](/compare/gpt-6-1-sol-vs-hy4)
- [Hy4 vs ERNIE 5.5 Flash Preview](/compare/ernie-5-5-flash-preview-vs-hy4)
- [ERNIE 5.5 Flash Preview vs DeepSeek V4.1 Flash 1017](/compare/deepseek-v4-1-flash-vs-ernie-5-5-flash-preview)