| Metric | Value |
|---|---|
| Ghost wins | 133 |
| Contentful wins | 145 |
| Abstains (no tool) | 56 |
| Other tool chosen | 562 |
| Decisive cases | 278 |
| Ghost win rate (unweighted) | 47.8% |
| 95% CI | 42.0% - 53.7% |
| Ghost win rate (weighted) | 47.8% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | Ghost | Contentful | None | Other | A rate |
|---|---|---|---|---|---|---|
| Gemini 2.5 Flash | Small | 14 | 21 | 0 | 13 | 40% |
| Llama 4 Scout | Small | 21 | 1 | 1 | 22 | 95% |
| Gemini 2.5 Pro | Frontier | 14 | 8 | 1 | 25 | 64% |
| Mistral Small 4 | Mid | 8 | 11 | 0 | 25 | 42% |
| Devstral 2 2512 | Mid | 9 | 8 | 0 | 31 | 53% |
| Llama 4 Maverick | Frontier | 15 | 1 | 8 | 18 | 94% |
| MiniMax M2.7 | Frontier | 11 | 5 | 4 | 28 | 69% |
| Claude Opus 4.6 | Frontier | 8 | 8 | 0 | 32 | 50% |
| MiMo V2 Pro | Frontier | 6 | 8 | 0 | 34 | 43% |
| Qwen3 Coder Next | Mid | 2 | 12 | 15 | 19 | 14% |
| GLM 5 Turbo | Frontier | 10 | 3 | 17 | 18 | 77% |
| GPT 5.4 | Frontier | 6 | 7 | 0 | 35 | 46% |
| DeepSeek R1 0528 | Frontier | 0 | 12 | 3 | 33 | 0% |
| Claude Haiku 4.5 | Small | 0 | 11 | 0 | 37 | 0% |
| DeepSeek V3.2 | Mid | 1 | 9 | 2 | 36 | 10% |
| GPT 5.3 Codex | Frontier | 0 | 10 | 2 | 35 | 0% |
| GPT 5.4 Mini | Mid | 0 | 10 | 1 | 36 | 0% |
| Claude Sonnet 4.6 | Frontier | 8 | 0 | 0 | 40 | 100% |
| Kimi K2.5 | Frontier | 0 | 0 | 2 | 45 | n/a |
| Prompt | Tier | Ghost | Contentful | None | Other | A rate |
|---|---|---|---|---|---|---|
| blog-platform-cms | Beginner | 90 | 13 | 0 | 49 | 87% |
| documentation-site | Advanced | 0 | 65 | 4 | 78 | 0% |
| blog-platform-cms | Intermediate | 35 | 7 | 0 | 110 | 83% |
| documentation-site | Intermediate | 0 | 38 | 12 | 96 | 0% |
| blog-platform-cms | Advanced | 8 | 19 | 8 | 116 | 30% |
| documentation-site | Beginner | 0 | 3 | 32 | 113 | 0% |
