| Metric | Value |
|---|---|
| LangSmith wins | 525 |
| Langfuse wins | 279 |
| Abstains (no tool) | 22 |
| Other tool chosen | 183 |
| Decisive cases | 804 |
| LangSmith win rate (unweighted) | 65.3% |
| 95% CI | 61.9% - 68.5% |
| LangSmith win rate (weighted) | 65.3% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | LangSmith | Langfuse | None | Other | A rate |
|---|---|---|---|---|---|---|
| MiMo V2 Pro | Frontier | 53 | 1 | 0 | 0 | 98% |
| GLM 5 Turbo | Frontier | 50 | 4 | 0 | 0 | 93% |
| Claude Sonnet 4.6 | Frontier | 37 | 17 | 0 | 0 | 69% |
| Claude Opus 4.6 | Frontier | 7 | 47 | 0 | 0 | 13% |
| GPT 5.3 Codex | Frontier | 6 | 48 | 0 | 0 | 11% |
| GPT 5.4 | Frontier | 3 | 51 | 0 | 0 | 6% |
| DeepSeek V3.2 | Mid | 49 | 3 | 0 | 2 | 94% |
| GPT 5.4 Mini | Mid | 41 | 10 | 1 | 2 | 80% |
| Gemini 2.5 Pro | Frontier | 49 | 1 | 4 | 0 | 98% |
| DeepSeek R1 0528 | Frontier | 48 | 0 | 1 | 5 | 100% |
| Claude Haiku 4.5 | Small | 14 | 34 | 0 | 6 | 29% |
| Kimi K2.5 | Frontier | 23 | 24 | 1 | 0 | 49% |
| Mistral Small 4 | Mid | 43 | 2 | 0 | 7 | 96% |
| Qwen3 Coder Next | Mid | 24 | 21 | 1 | 8 | 53% |
| MiniMax M2.7 | Frontier | 40 | 1 | 1 | 11 | 98% |
| Llama 4 Maverick | Frontier | 28 | 4 | 0 | 22 | 88% |
| Llama 4 Scout | Small | 0 | 11 | 5 | 35 | 0% |
| Devstral 2 2512 | Mid | 9 | 0 | 8 | 33 | 100% |
| Gemini 2.5 Flash | Small | 1 | 0 | 0 | 52 | 100% |
| Prompt | Tier | LangSmith | Langfuse | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-support-agent-platform | Intermediate | 107 | 34 | 1 | 29 | 76% |
| ai-revenue-ops-copilot | Intermediate | 98 | 37 | 0 | 29 | 73% |
| ai-support-agent-platform | Advanced | 86 | 49 | 0 | 36 | 64% |
| ai-support-agent-platform | Beginner | 70 | 64 | 5 | 30 | 52% |
| ai-revenue-ops-copilot | Advanced | 86 | 44 | 1 | 33 | 66% |
| ai-revenue-ops-copilot | Beginner | 78 | 51 | 15 | 26 | 60% |
