| Metric | Value |
|---|---|
| Langfuse wins | 279 |
| LangSmith wins | 525 |
| Abstains (no tool) | 22 |
| Other tool chosen | 183 |
| Decisive cases | 804 |
| Langfuse win rate (unweighted) | 34.7% |
| 95% CI | 31.5% - 38.1% |
| Langfuse win rate (weighted) | 34.7% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | Langfuse | LangSmith | None | Other | A rate |
|---|---|---|---|---|---|---|
| GPT 5.4 | Frontier | 51 | 3 | 0 | 0 | 94% |
| GPT 5.3 Codex | Frontier | 48 | 6 | 0 | 0 | 89% |
| Claude Opus 4.6 | Frontier | 47 | 7 | 0 | 0 | 87% |
| Claude Sonnet 4.6 | Frontier | 17 | 37 | 0 | 0 | 31% |
| GLM 5 Turbo | Frontier | 4 | 50 | 0 | 0 | 7% |
| MiMo V2 Pro | Frontier | 1 | 53 | 0 | 0 | 2% |
| DeepSeek V3.2 | Mid | 3 | 49 | 0 | 2 | 6% |
| GPT 5.4 Mini | Mid | 10 | 41 | 1 | 2 | 20% |
| Gemini 2.5 Pro | Frontier | 1 | 49 | 4 | 0 | 2% |
| Claude Haiku 4.5 | Small | 34 | 14 | 0 | 6 | 71% |
| DeepSeek R1 0528 | Frontier | 0 | 48 | 1 | 5 | 0% |
| Kimi K2.5 | Frontier | 24 | 23 | 1 | 0 | 51% |
| Qwen3 Coder Next | Mid | 21 | 24 | 1 | 8 | 47% |
| Mistral Small 4 | Mid | 2 | 43 | 0 | 7 | 4% |
| MiniMax M2.7 | Frontier | 1 | 40 | 1 | 11 | 2% |
| Llama 4 Maverick | Frontier | 4 | 28 | 0 | 22 | 13% |
| Llama 4 Scout | Small | 11 | 0 | 5 | 35 | 100% |
| Devstral 2 2512 | Mid | 0 | 9 | 8 | 33 | 0% |
| Gemini 2.5 Flash | Small | 0 | 1 | 0 | 52 | 0% |
| Prompt | Tier | Langfuse | LangSmith | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-support-agent-platform | Intermediate | 34 | 107 | 1 | 29 | 24% |
| ai-support-agent-platform | Advanced | 49 | 86 | 0 | 36 | 36% |
| ai-revenue-ops-copilot | Intermediate | 37 | 98 | 0 | 29 | 27% |
| ai-support-agent-platform | Beginner | 64 | 70 | 5 | 30 | 48% |
| ai-revenue-ops-copilot | Advanced | 44 | 86 | 1 | 33 | 34% |
| ai-revenue-ops-copilot | Beginner | 51 | 78 | 15 | 26 | 40% |
