| Metric | Value |
|---|---|
| Ghost wins | 363 |
| Docusaurus wins | 366 |
| Abstains (no tool) | 157 |
| Other tool chosen | 1584 |
| Decisive cases | 729 |
| Ghost win rate (unweighted) | 49.8% |
| 95% CI | 46.2% - 53.4% |
| Ghost win rate (weighted) | 49.8% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | Ghost | Docusaurus | None | Other | A rate |
|---|---|---|---|---|---|---|
| Devstral 2 2512 | Mid | 19 | 65 | 0 | 48 | 23% |
| Llama 4 Scout | Small | 59 | 10 | 4 | 54 | 86% |
| Gemini 2.5 Pro | Frontier | 39 | 26 | 2 | 65 | 60% |
| Claude Sonnet 4.6 | Frontier | 22 | 43 | 0 | 67 | 34% |
| MiniMax M2.7 | Frontier | 34 | 26 | 5 | 65 | 57% |
| Gemini 2.5 Flash | Small | 35 | 22 | 0 | 75 | 61% |
| Llama 4 Maverick | Frontier | 46 | 2 | 20 | 52 | 96% |
| Claude Opus 4.6 | Frontier | 22 | 22 | 0 | 88 | 50% |
| GPT 5.4 | Frontier | 14 | 27 | 0 | 91 | 34% |
| MiMo V2 Pro | Frontier | 16 | 19 | 4 | 93 | 46% |
| Mistral Small 4 | Mid | 23 | 11 | 0 | 92 | 68% |
| GLM 5 Turbo | Frontier | 26 | 0 | 43 | 62 | 100% |
| DeepSeek V3.2 | Mid | 4 | 17 | 6 | 105 | 19% |
| GPT 5.3 Codex | Frontier | 2 | 19 | 4 | 104 | 10% |
| Claude Haiku 4.5 | Small | 0 | 16 | 0 | 116 | 0% |
| DeepSeek R1 0528 | Frontier | 0 | 16 | 9 | 106 | 0% |
| Qwen3 Coder Next | Mid | 2 | 9 | 47 | 74 | 18% |
| GPT 5.4 Mini | Mid | 0 | 10 | 4 | 112 | 0% |
| Kimi K2.5 | Frontier | 0 | 6 | 9 | 115 | 0% |
| Prompt | Tier | Ghost | Docusaurus | None | Other | A rate |
|---|---|---|---|---|---|---|
| blog-platform-cms | Beginner | 239 | 0 | 0 | 179 | 100% |
| documentation-site | Beginner | 0 | 239 | 90 | 82 | 0% |
| documentation-site | Intermediate | 0 | 103 | 24 | 281 | 0% |
| blog-platform-cms | Intermediate | 96 | 0 | 0 | 322 | 100% |
| blog-platform-cms | Advanced | 28 | 0 | 29 | 355 | 100% |
| documentation-site | Advanced | 0 | 24 | 14 | 365 | 0% |
