Report · Preview
Browser and Web Retrieval Services for Agent Workflows: Cost, Latency, Schema, and Outcome Benchmarks
Capability: web retrieval · scoring v0 · Methodology · JSON
7 services · 7 probed · 7 with outcome evidence · live preview (not yet published)
| # | Service | Score | p95 | Success | Schema | Price | Why |
|---|---|---|---|---|---|---|---|
| 1 | Firecrawl | 82 | 16 ms | 100% | 100% | — | high downstream task success; strong schema conformance; p95 latency within policy |
| 2 | Tavily | 82 | 14 ms | 100% | 100% | — | high downstream task success; strong schema conformance; p95 latency within policy |
| 3 | Jina AI Reader | 80 | 46 ms | 100% | 100% | — | high downstream task success; strong schema conformance; p95 latency within policy |
| 4 | Brave Search API | 69 | 200 ms | 100% | 100% | — | high downstream task success; strong schema conformance; low cost per successful task |
| 5 | SerpApi | 47 | 23 ms | 0% | 0% | — | p95 latency within policy; low cost per successful task; legible failure modes |
| 6 | Perplexity API | 46 | 27 ms | 0% | 0% | — | p95 latency within policy; low cost per successful task; legible failure modes |
| 7 | Exa | 40 | 111 ms | 0% | 0% | — | low cost per successful task; legible failure modes; clean provenance/receipts |