| Metric | Value |
|---|---|
| Braintrust wins | 303 |
| LangSmith wins | 330 |
| Abstains (no tool) | 36 |
| Other tool chosen | 331 |
| Decisive cases | 633 |
| Braintrust win rate (unweighted) | 47.9% |
| 95% CI | 44.0% - 51.8% |
| Braintrust win rate (weighted) | 47.9% |
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 | 51 | 3 | 0 | 0 | 94% |
| Claude Haiku 4.5 | Small | 44 | 3 | 1 | 4 | 94% |
| Claude Opus 4.6 | Frontier | 46 | 0 | 0 | 8 | 100% |
| GPT 5.4 | Frontier | 39 | 7 | 0 | 8 | 85% |
| Claude Sonnet 4.6 | Frontier | 28 | 18 | 0 | 8 | 61% |
| Kimi K2.5 | Frontier | 44 | 0 | 0 | 4 | 100% |
| DeepSeek R1 0528 | Frontier | 0 | 44 | 2 | 8 | 0% |
| GLM 5 Turbo | Frontier | 32 | 10 | 7 | 5 | 76% |
| Gemini 2.5 Pro | Frontier | 0 | 42 | 6 | 6 | 0% |
| MiniMax M2.7 | Frontier | 17 | 23 | 1 | 11 | 43% |
| Mistral Small 4 | Mid | 0 | 40 | 0 | 11 | 0% |
| GPT 5.4 Mini | Mid | 1 | 36 | 1 | 16 | 3% |
| Qwen3 Coder Next | Mid | 0 | 35 | 3 | 16 | 0% |
| DeepSeek V3.2 | Mid | 0 | 32 | 9 | 11 | 0% |
| MiMo V2 Pro | Frontier | 1 | 26 | 2 | 25 | 4% |
| Llama 4 Maverick | Frontier | 0 | 8 | 0 | 45 | 0% |
| Devstral 2 2512 | Mid | 0 | 2 | 1 | 48 | 0% |
| Gemini 2.5 Flash | Small | 0 | 1 | 0 | 51 | 0% |
| Llama 4 Scout | Small | 0 | 0 | 3 | 46 | n/a |
| Prompt | Tier | Braintrust | LangSmith | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-revenue-ops-copilot | Intermediate | 49 | 73 | 1 | 41 | 40% |
| ai-support-agent-platform | Advanced | 52 | 60 | 1 | 56 | 46% |
| ai-support-agent-platform | Intermediate | 37 | 74 | 4 | 51 | 33% |
| ai-revenue-ops-copilot | Beginner | 63 | 47 | 4 | 54 | 57% |
| ai-revenue-ops-copilot | Advanced | 63 | 45 | 1 | 55 | 58% |
| ai-support-agent-platform | Beginner | 39 | 31 | 25 | 74 | 56% |
