| Metric | Value |
|---|---|
| Tavily wins | 372 |
| SerpAPI wins | 57 |
| Abstains (no tool) | 1 |
| Other tool chosen | 62 |
| Decisive cases | 429 |
| Tavily win rate (unweighted) | 86.7% |
| 95% CI | 83.2% - 89.6% |
| Tavily win rate (weighted) | 86.7% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model range | Tier | Tavily | SerpAPI | None | Other | A rate |
|---|---|---|---|---|---|---|
| DeepSeek Flash 3.2–4.1 | Mid | 27 | 0 | 0 | 0 | 100% |
| DeepSeek Pro 4–4 0813 | Frontier | 25 | 1 | 0 | 1 | 96% |
| DeepSeek R1 | Frontier | 0 | 22 | 1 | 4 | 0% |
| Devstral | Mid | 5 | 13 | 0 | 0 | 28% |
| Gemini Flash 2.5–3.8 | Small | 24 | 0 | 0 | 0 | 100% |
| Gemini Pro 2.5–3.1 Preview | Frontier | 6 | 0 | 0 | 0 | 100% |
| GLM 5 Turbo–5.3 | Frontier | 27 | 0 | 0 | 0 | 100% |
| GPT 5.4–6 Astra | Frontier | 27 | 0 | 0 | 0 | 100% |
| GPT Coding 5.3 Codex–6.1 Sol | Frontier | 27 | 0 | 0 | 0 | 100% |
| GPT Mini 5.4 Mini–6 Luna | Mid | 26 | 0 | 0 | 1 | 100% |
| Haiku | Small | 15 | 6 | 0 | 6 | 71% |
| Kimi 2.5–3 | Frontier | 27 | 0 | 0 | 0 | 100% |
| Llama Maverick | Frontier | 2 | 1 | 0 | 22 | 67% |
| Llama Scout | Small | 0 | 1 | 0 | 23 | 0% |
| MiMo Pro 2–2.6 | Frontier | 21 | 1 | 0 | 4 | 95% |
| MiniMax M 2.7–3 | Frontier | 27 | 0 | 0 | 0 | 100% |
| Mistral Small | Mid | 15 | 12 | 0 | 0 | 56% |
| Opus 4.6–5.5 | Frontier | 27 | 0 | 0 | 0 | 100% |
| Qwen Coder | Mid | 17 | 0 | 0 | 1 | 100% |
| Sonnet 4.6–5.5 | Frontier | 27 | 0 | 0 | 0 | 100% |
| Prompt | Tier | Tavily | SerpAPI | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-agent-application | Advanced | 129 | 16 | 0 | 15 | 89% |
| ai-agent-application | Intermediate | 117 | 26 | 1 | 22 | 82% |
| ai-agent-application | Beginner | 126 | 15 | 0 | 25 | 89% |
