| Metric | Value |
|---|---|
| Braintrust wins | 720 |
| LangSmith wins | 837 |
| Abstains (no tool) | 90 |
| Other tool chosen | 797 |
| Decisive cases | 1557 |
| Braintrust win rate (unweighted) | 46.2% |
| 95% CI | 43.8% - 48.7% |
| Braintrust win rate (weighted) | 46.2% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | Braintrust | LangSmith | None | Other | A rate |
|---|---|---|---|---|---|---|
| GPT 5.3 Codex | Frontier | 126 | 6 | 0 | 0 | 95% |
| Claude Haiku 4.5 | Small | 103 | 13 | 1 | 8 | 89% |
| Claude Opus 4.6 | Frontier | 113 | 1 | 0 | 18 | 99% |
| Claude Sonnet 4.6 | Frontier | 58 | 55 | 0 | 19 | 51% |
| GPT 5.4 | Frontier | 92 | 18 | 0 | 22 | 84% |
| Kimi K2.5 | Frontier | 109 | 0 | 3 | 7 | 100% |
| DeepSeek R1 0528 | Frontier | 0 | 108 | 7 | 17 | 0% |
| Gemini 2.5 Pro | Frontier | 0 | 106 | 9 | 17 | 0% |
| GLM 5 Turbo | Frontier | 84 | 18 | 19 | 11 | 82% |
| GPT 5.4 Mini | Mid | 3 | 97 | 3 | 29 | 3% |
| Mistral Small 4 | Mid | 0 | 94 | 1 | 29 | 0% |
| MiniMax M2.7 | Frontier | 31 | 62 | 5 | 31 | 33% |
| Qwen3 Coder Next | Mid | 0 | 88 | 3 | 40 | 0% |
| DeepSeek V3.2 | Mid | 0 | 80 | 22 | 26 | 0% |
| MiMo V2 Pro | Frontier | 1 | 62 | 8 | 61 | 2% |
| Llama 4 Maverick | Frontier | 0 | 21 | 0 | 106 | 0% |
| Devstral 2 2512 | Mid | 0 | 5 | 4 | 116 | 0% |
| Gemini 2.5 Flash | Small | 0 | 3 | 1 | 123 | 0% |
| Llama 4 Scout | Small | 0 | 0 | 4 | 117 | n/a |
| Prompt | Tier | Braintrust | LangSmith | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-revenue-ops-copilot | Intermediate | 122 | 182 | 4 | 96 | 40% |
| ai-support-agent-platform | Intermediate | 87 | 195 | 5 | 122 | 31% |
| ai-revenue-ops-copilot | Advanced | 153 | 124 | 2 | 121 | 55% |
| ai-revenue-ops-copilot | Beginner | 143 | 119 | 10 | 138 | 55% |
| ai-support-agent-platform | Advanced | 123 | 134 | 5 | 148 | 48% |
| ai-support-agent-platform | Beginner | 92 | 83 | 64 | 172 | 53% |
