| Metric | Value |
|---|---|
| LangSmith wins | 837 |
| Braintrust wins | 720 |
| Abstains (no tool) | 90 |
| Other tool chosen | 797 |
| Decisive cases | 1557 |
| LangSmith win rate (unweighted) | 53.8% |
| 95% CI | 51.3% - 56.2% |
| LangSmith win rate (weighted) | 53.8% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | LangSmith | Braintrust | None | Other | A rate |
|---|---|---|---|---|---|---|
| GPT 5.3 Codex | Frontier | 6 | 126 | 0 | 0 | 5% |
| Claude Haiku 4.5 | Small | 13 | 103 | 1 | 8 | 11% |
| Claude Opus 4.6 | Frontier | 1 | 113 | 0 | 18 | 1% |
| Claude Sonnet 4.6 | Frontier | 55 | 58 | 0 | 19 | 49% |
| GPT 5.4 | Frontier | 18 | 92 | 0 | 22 | 16% |
| Kimi K2.5 | Frontier | 0 | 109 | 3 | 7 | 0% |
| DeepSeek R1 0528 | Frontier | 108 | 0 | 7 | 17 | 100% |
| Gemini 2.5 Pro | Frontier | 106 | 0 | 9 | 17 | 100% |
| GLM 5 Turbo | Frontier | 18 | 84 | 19 | 11 | 18% |
| GPT 5.4 Mini | Mid | 97 | 3 | 3 | 29 | 97% |
| Mistral Small 4 | Mid | 94 | 0 | 1 | 29 | 100% |
| MiniMax M2.7 | Frontier | 62 | 31 | 5 | 31 | 67% |
| Qwen3 Coder Next | Mid | 88 | 0 | 3 | 40 | 100% |
| DeepSeek V3.2 | Mid | 80 | 0 | 22 | 26 | 100% |
| MiMo V2 Pro | Frontier | 62 | 1 | 8 | 61 | 98% |
| Llama 4 Maverick | Frontier | 21 | 0 | 0 | 106 | 100% |
| Devstral 2 2512 | Mid | 5 | 0 | 4 | 116 | 100% |
| Gemini 2.5 Flash | Small | 3 | 0 | 1 | 123 | 100% |
| Llama 4 Scout | Small | 0 | 0 | 4 | 117 | n/a |
| Prompt | Tier | LangSmith | Braintrust | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-revenue-ops-copilot | Intermediate | 182 | 122 | 4 | 96 | 60% |
| ai-support-agent-platform | Intermediate | 195 | 87 | 5 | 122 | 69% |
| ai-revenue-ops-copilot | Advanced | 124 | 153 | 2 | 121 | 45% |
| ai-revenue-ops-copilot | Beginner | 119 | 143 | 10 | 138 | 45% |
| ai-support-agent-platform | Advanced | 134 | 123 | 5 | 148 | 52% |
| ai-support-agent-platform | Beginner | 83 | 92 | 64 | 172 | 47% |
