| Metric | Value |
|---|---|
| Stream wins | 224 |
| Pusher wins | 497 |
| Abstains (no tool) | 40 |
| Other tool chosen | 2636 |
| Decisive cases | 721 |
| Stream win rate (unweighted) | 31.1% |
| 95% CI | 27.8% - 34.5% |
| Stream win rate (weighted) | 31.1% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | Stream | Pusher | None | Other | A rate |
|---|---|---|---|---|---|---|
| GLM 5 Turbo | Frontier | 23 | 68 | 3 | 86 | 25% |
| Gemini 2.5 Pro | Frontier | 6 | 79 | 0 | 95 | 7% |
| MiMo V2 Pro | Frontier | 4 | 76 | 0 | 99 | 5% |
| Qwen3 Coder Next | Mid | 0 | 74 | 0 | 105 | 0% |
| Kimi K2.5 | Frontier | 49 | 2 | 5 | 122 | 96% |
| Claude Opus 4.6 | Frontier | 21 | 27 | 0 | 131 | 44% |
| GPT 5.4 | Frontier | 32 | 14 | 0 | 131 | 70% |
| Gemini 2.5 Flash | Small | 28 | 7 | 0 | 145 | 80% |
| Llama 4 Maverick | Frontier | 9 | 23 | 0 | 148 | 28% |
| DeepSeek R1 0528 | Frontier | 2 | 30 | 30 | 117 | 6% |
| Llama 4 Scout | Small | 0 | 31 | 0 | 147 | 0% |
| Mistral Small 4 | Mid | 0 | 30 | 0 | 149 | 0% |
| Claude Sonnet 4.6 | Frontier | 21 | 7 | 0 | 152 | 75% |
| GPT 5.4 Mini | Mid | 15 | 5 | 2 | 154 | 75% |
| MiniMax M2.7 | Frontier | 6 | 10 | 0 | 162 | 38% |
| DeepSeek V3.2 | Mid | 5 | 9 | 0 | 166 | 36% |
| GPT 5.3 Codex | Frontier | 3 | 2 | 0 | 175 | 60% |
| Devstral 2 2512 | Mid | 0 | 2 | 0 | 178 | 0% |
| Claude Haiku 4.5 | Small | 0 | 1 | 0 | 174 | 0% |
| Prompt | Tier | Stream | Pusher | None | Other | A rate |
|---|---|---|---|---|---|---|
| chat-application | Beginner | 159 | 12 | 3 | 205 | 93% |
| social-media-platform | Intermediate | 7 | 115 | 4 | 251 | 6% |
| social-media-platform | Advanced | 7 | 99 | 3 | 270 | 7% |
| social-media-platform | Beginner | 11 | 91 | 17 | 254 | 11% |
| chat-application | Advanced | 17 | 41 | 1 | 317 | 29% |
| project-management-tool | Beginner | 0 | 53 | 4 | 322 | 0% |
| project-management-tool | Intermediate | 0 | 47 | 3 | 327 | 0% |
| chat-application | Intermediate | 22 | 12 | 5 | 339 | 65% |
| project-management-tool | Advanced | 1 | 27 | 0 | 351 | 4% |
