{ "schema_version": 1, "source": "https://nativeport.ai/evals.json", "latest_run": "2026-06-22", "methodology": "One fixed task corpus per capability, identical for every provider; scorecards carry their run dates. Weak scores stay published, and the gateway's flat top-up fee means the ranking earns nothing from steering you toward pricier providers. Details: https://nativeport.ai/methodology/", "gateway": { "base_url": "https://api.nativeport.ai", "auth": "Authorization: Bearer ", "route_rule": "/{provider}/{provider-native-path} — forwarded to the provider's native API, response returned unmodified", "empty_balance": "HTTP 402 until credits are added (deterministic; optional auto-top-up)", "sign_up": "https://accounts.nativeport.ai/sign-up", "panel": "https://panel.nativeport.ai", "docs": "https://docs.nativeport.ai", "pricing": "https://nativeport.ai/pricing/" }, "verbs": { "search": { "label": "Search", "description": "Web search: a query in, ranked result links out" }, "scrape": { "label": "Scrape", "description": "Fetch one URL as clean content/markdown, anti-bot handled" }, "crawl": { "label": "Crawl", "description": "Enumerate and fetch many pages across a site" }, "extract_ai": { "label": "Extract · AI/schema", "description": "Structured fields pulled from a page against an AI/natural-language schema" }, "extract_rules": { "label": "Extract · CSS rules", "description": "Structured fields pulled from a page via CSS/XPath rules" }, "answer": { "label": "Answer", "description": "A question in, a synthesized answer with cited sources out" }, "screenshot": { "label": "Screenshot", "description": "Rendered image of a URL" }, "scrape_domain": { "label": "Scrape-domain", "description": "Purpose-built scrapers for hard, high-value domains (e-commerce, social)" }, "act": { "label": "Act · declarative", "description": "Declarative in-page interactions (click, fill, scroll) in one call" }, "act_agent": { "label": "Act · NL-agent", "description": "A natural-language goal handed to an autonomous browser agent" }, "serp": { "label": "SERP verticals", "description": "Google SERP verticals (web, news, images, places, scholar) as structured JSON" }, "parse": { "label": "Parse · PDF/doc", "description": "PDF/document to text, with OCR where the provider supports it" }, "watch": { "label": "Watch", "description": "Change detection / monitoring on a URL" } }, "providers": { "serper": { "name": "Serper", "group": "Search", "category": "Google SERP scrape", "tier": "T1", "route": "/serper", "call": "https://api.nativeport.ai/serper/", "docs": "https://docs.nativeport.ai/serper", "page": "https://nativeport.ai/providers/serper/", "pricing_entry": "$1.00 → $0.30 / 1k", "latency_note": "~0.6–1.7s", "provider_mcp": "community", "summary": "Google search results across web, news, images, places, maps and scholar verticals, plus an endpoint for scraping webpages.", "choose_if": "Pick it when the goal is the lowest-cost, fastest Google grounding and you'll synthesize the results on your own.", "avoid_if": "Not the pick if you require coverage beyond Google, richer SERP elements, or content extraction out of the box.", "evals": [ { "verb": "serp", "label": "SERP verticals", "composite": 9.56, "rank": 1, "of": 4, "top": true, "measured": "2026-06-22", "note": "Lowest cost and latency; Google only.", "metrics": [ { "key": "quality", "label": "Quality", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1003.8, "value": "1.0 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.0003, "value": "$0.0003 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 7.24, "rank": 3, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.569, "value": "0.57" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1480.8, "value": "1.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.000527, "value": "$0.00053 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "serpapi": { "name": "SerpApi", "group": "Search", "category": "Multi-engine SERP scrape", "tier": "T1", "route": "/serpapi", "call": "https://api.nativeport.ai/serpapi/", "docs": "https://docs.nativeport.ai/serpapi", "page": "https://nativeport.ai/providers/serpapi/", "pricing_entry": "~$15 → $9 / 1k", "latency_note": "~1.7–2.0s", "provider_mcp": "official", "summary": "Google plus 80+ additional search engines as structured results — key handling is taken off your plate.", "choose_if": "The right call for compliance-sensitive teams, for breadth across many engines, or as the drop-in default in LangChain.", "avoid_if": "Skip it when queries are Google-only and price matters — Serper undercuts it there.", "evals": [ { "verb": "serp", "label": "SERP verticals", "composite": 9.0, "rank": 3, "of": 4, "top": false, "measured": "2026-06-22", "note": "Every tested vertical is covered.", "metrics": [ { "key": "quality", "label": "Quality", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1424.3, "value": "1.4 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.015, "value": "$0.015 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 6.1, "rank": 9, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.588, "value": "0.59" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2826.2, "value": "2.8 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.015307, "value": "$0.015 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "searchapi": { "name": "SearchAPI.io", "group": "Search", "category": "Multi-engine SERP scrape", "tier": "T2", "route": "/searchapi", "call": "https://api.nativeport.ai/searchapi/", "docs": "https://docs.nativeport.ai/searchapi", "page": "https://nativeport.ai/providers/searchapi/", "pricing_entry": "$1.00–$4.00 / 1k", "latency_note": "real-time", "provider_mcp": "official", "summary": "One endpoint serving search results from 80-plus engines, Google, Bing and YouTube among them.", "choose_if": "Reach for it if you're after SerpApi-compatible JSON, cheaper, and billed only per successful search.", "avoid_if": "The catch: since Sept-2025 the num parameter is silently capped at 10, so large Google result pages are out.", "evals": [ { "verb": "serp", "label": "SERP verticals", "composite": 9.02, "rank": 2, "of": 4, "top": false, "measured": "2026-06-22", "note": "Widest vertical coverage, with SerpApi-compatible JSON.", "metrics": [ { "key": "quality", "label": "Quality", "raw": 0.985, "value": "0.98" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1930.0, "value": "1.9 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.004, "value": "$0.004 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 6.44, "rank": 8, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.574, "value": "0.57" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2485.2, "value": "2.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.006968, "value": "$0.007 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "brave": { "name": "Brave Search", "group": "Search", "category": "Independent crawler index", "tier": "T1", "route": "/brave", "call": "https://api.nativeport.ai/brave/", "docs": "https://docs.nativeport.ai/brave", "page": "https://nativeport.ai/providers/brave/", "pricing_entry": "$5 / 1k", "latency_note": null, "provider_mcp": "official", "summary": "A 30B+ page search index built independently, spanning news, video, image and web verticals.", "choose_if": "Pick it when independence from Google and Bing matters, along with a privacy posture and grounding for AI.", "avoid_if": "Skip it when a single query has to return more than roughly 200 results.", "evals": [ { "verb": "answer", "label": "Answer", "composite": 7.02, "rank": 5, "of": 7, "top": false, "measured": "2026-06-21", "note": "Accurate answers, though per-query cost far exceeds peers.", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.279, "value": "0.28" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2388.4, "value": "2.4 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.06, "value": "$0.06 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 6.99, "rank": 4, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.639, "value": "0.64" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1271.4, "value": "1.3 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.007826, "value": "$0.0078 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "youcom": { "name": "You.com", "group": "Search", "category": "Own crawl index + Research API", "tier": "T2", "route": "/youcom", "call": "https://api.nativeport.ai/youcom/", "docs": "https://docs.nativeport.ai/youcom", "page": "https://nativeport.ai/providers/youcom/", "pricing_entry": "$5 / 1k", "latency_note": null, "provider_mcp": "official", "summary": "Own-index web search with an accompanying Research API that cites sources, aimed at grounding LLMs.", "choose_if": "For teams that want soft domain-boosting on an independent index, with a Research API that returns cited answers.", "avoid_if": "Not the pick if a flat response shape matters: results sit nested in results.web[] with snippets[], so an adapter is required.", "evals": [ { "verb": "answer", "label": "Answer", "composite": 7.79, "rank": 2, "of": 7, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.51, "value": "0.51" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4341.7, "value": "4.3 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.02, "value": "$0.02 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 7.41, "rank": 2, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.653, "value": "0.65" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1357.0, "value": "1.4 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.00766, "value": "$0.0077 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "dataforseo": { "name": "DataForSEO", "group": "Search", "category": "SEO & SERP data platform", "tier": "T2", "route": "/dataforseo", "call": "https://api.nativeport.ai/dataforseo/", "docs": "https://docs.nativeport.ai/dataforseo", "page": "https://nativeport.ai/providers/dataforseo/", "pricing_entry": "$0.60 async / $2 live / 1k", "latency_note": "~5s std / ~6s live", "provider_mcp": "official", "summary": "SERP and SEO intelligence — keywords, rankings, backlinks — with the deepest taxonomy of SERP elements.", "choose_if": "Reach for it if the job is bulk asynchronous SERP scraping with webhooks and the fullest SERP-feature taxonomy.", "avoid_if": "The catch: responses nest deeply (tasks[].result[].items[]) and latency isn't low, which is genuine friction.", "evals": [ { "verb": "serp", "label": "SERP verticals", "composite": 8.09, "rank": 4, "of": 4, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "quality", "label": "Quality", "raw": 0.929, "value": "0.93" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 8214.0, "value": "8.2 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.002, "value": "$0.002 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 7.1, "value": "7%" } ] }, { "verb": "search", "label": "Search", "composite": 5.86, "rank": 10, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.588, "value": "0.59" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2903.3, "value": "2.9 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.003402, "value": "$0.0034 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "tavily": { "name": "Tavily", "group": "Search", "category": "Agent-native search index", "tier": "T1", "route": "/tavily", "call": "https://api.nativeport.ai/tavily/", "docs": "https://docs.nativeport.ai/tavily", "page": "https://nativeport.ai/providers/tavily/", "pricing_entry": "$8 (basic) / $16 (adv) / 1k", "latency_note": "~1.0s", "provider_mcp": "official", "summary": "A search index native to agents and RAG, returning snippets with relevance scores and an optional answer.", "choose_if": "Pick it when you're building the typical agent or RAG stack and want relevance scores plus an optional answer from a single call.", "avoid_if": "Not the pick if per-result dates matter, or if a documented SLA/compliance posture is required.", "evals": [ { "verb": "crawl", "label": "Crawl", "composite": 8.38, "rank": 3, "of": 5, "top": false, "measured": "2026-06-22", "note": "Performs well through its /map variant.", "metrics": [ { "key": "coverage", "label": "Coverage", "raw": 0.78, "value": "0.78" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 174.1, "value": "174 ms" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.013846, "value": "$0.014 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "answer", "label": "Answer", "composite": 7.13, "rank": 4, "of": 7, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.243, "value": "0.24" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1375.6, "value": "1.4 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.008, "value": "$0.008 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 6.66, "rank": 7, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.583, "value": "0.58" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2094.1, "value": "2.1 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.013714, "value": "$0.014 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "exa": { "name": "Exa", "group": "Search", "category": "Neural + keyword index", "tier": "T1", "route": "/exa", "call": "https://api.nativeport.ai/exa/", "docs": "https://docs.nativeport.ai/exa", "page": "https://nativeport.ai/providers/exa/", "pricing_entry": "$7 / 1k (+content $1/1k)", "latency_note": null, "provider_mcp": "official", "summary": "Semantic search driven by neural embeddings, offering find-similar, clean extraction and an integrated answer endpoint.", "choose_if": "Reach for it if the task is semantic discovery, find-similar lookups, research, or using its built-in /answer endpoint.", "avoid_if": "Skip it when full keyword/SERP parity is required; content and summary are billed per page and per type, and those charges stack.", "evals": [ { "verb": "answer", "label": "Answer", "composite": 7.9, "rank": 1, "of": 7, "top": true, "measured": "2026-06-21", "note": "Citation faithfulness ranks highest.", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.485, "value": "0.48" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1514.9, "value": "1.5 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.005, "value": "$0.005 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 6.7, "rank": 6, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.606, "value": "0.61" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1597.6, "value": "1.6 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.014015, "value": "$0.014 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "linkup": { "name": "Linkup", "group": "Search", "category": "Own crawl index", "tier": "T2", "route": "/linkup", "call": "https://api.nativeport.ai/linkup/", "docs": "https://docs.nativeport.ai/linkup", "page": "https://nativeport.ai/providers/linkup/", "pricing_entry": "$5 / 1k", "latency_note": null, "provider_mcp": "official", "summary": "A search alternative that stays ToS-clean, offering sourced answers natively, structured JSON output, and markdown via /fetch.", "choose_if": "The right call when compliance matters and you want a SERP alternative that returns sourced answers and structured JSON natively.", "avoid_if": "The catch: no relevance scores, no per-result dates, and no geo controls are exposed.", "evals": [ { "verb": "answer", "label": "Answer", "composite": 6.91, "rank": 6, "of": 7, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.16, "value": "0.16" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2520.3, "value": "2.5 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.005, "value": "$0.005 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 5.47, "rank": 11, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.394, "value": "0.39" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 3246.3, "value": "3.2 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.012706, "value": "$0.013 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "parallel": { "name": "Parallel", "group": "Search", "category": "Agent-optimized search index", "tier": "T2", "route": "/parallel", "call": "https://api.nativeport.ai/parallel/", "docs": "https://docs.nativeport.ai/parallel", "page": "https://nativeport.ai/providers/parallel/", "pricing_entry": "$5 (base) / $9 (pro) / 1k", "latency_note": "P50 <1s", "provider_mcp": "official + no-key MCP", "summary": "An agent-optimized index where the ranking follows a natural-language objective; results arrive as pre-compressed excerpts.", "choose_if": "Pick it for the least in-loop friction: a Search MCP that needs no key, with P50 under a second.", "avoid_if": "Not the pick if the Search response itself must include full content, relevance scores, or an answer already synthesized.", "evals": [ { "verb": "answer", "label": "Answer", "composite": 7.58, "rank": 3, "of": 7, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.485, "value": "0.48" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 16300.9, "value": "16.3 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.01, "value": "$0.01 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 6.76, "rank": 5, "of": 11, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.644, "value": "0.64" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2654.4, "value": "2.7 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.006216, "value": "$0.0062 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "jina": { "name": "Jina", "group": "Search", "category": "Search + read-to-markdown", "tier": "T2", "route": "/jina", "call": "https://api.nativeport.ai/jina/", "docs": "https://docs.nativeport.ai/jina", "page": "https://nativeport.ai/providers/jina/", "pricing_entry": "~token-based", "latency_note": "low RPM", "provider_mcp": "official", "summary": "A URL-to-markdown Reader that benchmarks at the top of its class, plus search that hands back full-page markdown in a single call.", "choose_if": "Reach for it if a URL needs to become clean, LLM-ready markdown, or you want search and read combined in one call.", "avoid_if": "Skip it when high search RPM or lots of results per query matter — the default is ≤5 results and RPM runs low.", "evals": [ { "verb": "parse", "label": "Parse · PDF/doc", "composite": 8.43, "rank": 2, "of": 3, "top": false, "measured": "2026-06-22", "note": "No OCR: scanned PDFs come back empty.", "metrics": [ { "key": "text_accuracy", "label": "Text accuracy", "raw": 0.78, "value": "0.78" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 262.8, "value": "263 ms" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.0, "value": "$0 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "search", "label": "Search", "composite": 7.69, "rank": 1, "of": 11, "top": true, "measured": "2026-06-20", "note": "First across all three query classes.", "metrics": [ { "key": "recall_at_10", "label": "Recall@10", "raw": 0.602, "value": "0.60" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1392.5, "value": "1.4 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.000332, "value": "$0.00033 / useful result" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "answer", "label": "Answer", "composite": 6.61, "rank": 7, "of": 7, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "answer_correctness", "label": "Correctness", "raw": 1.0, "value": "1.00" }, { "key": "citation_faithfulness", "label": "Citation faithfulness", "raw": 0.108, "value": "0.11" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2619.0, "value": "2.6 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.02, "value": "$0.02 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "scraperapi": { "name": "ScraperAPI", "group": "Scraping & Crawling", "category": "Single-call scraping API", "tier": "T1", "route": "/scraperapi", "call": "https://api.nativeport.ai/scraperapi/", "docs": "https://docs.nativeport.ai/scraperapi", "page": "https://nativeport.ai/providers/scraperapi/", "pricing_entry": "~$0.49 / 1k (credit ×mult 1→75)", "latency_note": null, "provider_mcp": "none", "summary": "General scraping behind a single GET, with proxy rotation and automatic retries handled.", "choose_if": "The right call when the simplest unblocking request wins and MCP-native discovery isn't on your list.", "avoid_if": "The catch: credit multipliers make costs hard to see upfront, and there's no first-party MCP.", "evals": [ { "verb": "screenshot", "label": "Screenshot", "composite": 7.78, "rank": 5, "of": 8, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 0.875, "value": "0.88" }, { "key": "full_page_support", "label": "Full-page", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 27445.9, "value": "27.4 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.0049, "value": "$0.0049 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 12.5, "value": "12%" } ] }, { "verb": "scrape_domain", "label": "Scrape-domain", "composite": 7.26, "rank": 4, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "value_accuracy", "label": "Accuracy", "raw": 0.733, "value": "0.73" }, { "key": "field_fill", "label": "Field fill", "raw": 0.778, "value": "0.78" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2270.2, "value": "2.3 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.006685, "value": "$0.0067 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "scrape", "label": "Scrape", "composite": 6.53, "rank": 10, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.625, "value": "62%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 4.4, "value": "4.4 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2360.6, "value": "2.4 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.010192, "value": "$0.01 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 37.5, "value": "38%" } ] } ] }, "firecrawl": { "name": "Firecrawl", "group": "Scraping & Crawling", "category": "Scrape + crawl + extract", "tier": "T1", "route": "/firecrawl", "call": "https://api.nativeport.ai/firecrawl/", "docs": "https://docs.nativeport.ai/firecrawl", "page": "https://nativeport.ai/providers/firecrawl/", "pricing_entry": "~$0.83 / 1k (credit/page)", "latency_note": null, "provider_mcp": "official", "summary": "LLM-ready markdown that tops the benchmarks, plus schema-driven extraction, asynchronous crawling and an official MCP.", "choose_if": "Pick it for the top-scoring scrape and markdown quality, with crawl, scrape, screenshot and extract on one surface.", "avoid_if": "Skip it when hard targets meet a tight budget: stealth plus JSON output stacks to about 5–9 times the base credit.", "evals": [ { "verb": "scrape", "label": "Scrape", "composite": 9.81, "rank": 1, "of": 10, "top": true, "measured": "2026-06-20", "note": "Class-leading result; every anti-bot vendor cleared.", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 1.0, "value": "100%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 9.62, "value": "9.6 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 962.2, "value": "962 ms" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.002094, "value": "$0.0021 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "screenshot", "label": "Screenshot", "composite": 9.5, "rank": 1, "of": 8, "top": true, "measured": "2026-06-21", "note": "Highest rate of valid images.", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 1.0, "value": "1.00" }, { "key": "full_page_support", "label": "Full-page", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1258.3, "value": "1.3 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.001, "value": "$0.001 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "parse", "label": "Parse · PDF/doc", "composite": 8.82, "rank": 1, "of": 3, "top": true, "measured": "2026-06-22", "note": "Scanned-PDF OCR sets it apart.", "metrics": [ { "key": "text_accuracy", "label": "Text accuracy", "raw": 0.907, "value": "0.91" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 293.2, "value": "293 ms" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.001103, "value": "$0.0011 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "extract_ai", "label": "Extract · AI/schema", "composite": 8.81, "rank": 2, "of": 4, "top": false, "measured": "2026-06-21", "note": "On the precision test it hallucinated a null field.", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 0.957, "value": "0.96" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 1795.9, "value": "1.8 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.005225, "value": "$0.0052 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "crawl", "label": "Crawl", "composite": 8.61, "rank": 1, "of": 5, "top": true, "measured": "2026-06-22", "note": "Leads for clean, path-scoped enumeration.", "metrics": [ { "key": "coverage", "label": "Coverage", "raw": 0.92, "value": "0.92" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 16260.6, "value": "16.3 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.023913, "value": "$0.024 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "watch", "label": "Watch", "composite": 8.33, "rank": 1, "of": 1, "top": true, "measured": "2026-06-21", "note": "Uniquely offers synchronous change detection.", "metrics": [ { "key": "classification_accuracy", "label": "Class. accuracy", "raw": 1.0, "value": "1.00" }, { "key": "diff_quality", "label": "Diff quality", "raw": 0.5, "value": "0.50" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2987.6, "value": "3.0 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.001, "value": "$0.001 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "act", "label": "Act · declarative", "composite": 7.41, "rank": 3, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 0.8, "value": "0.80" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4845.8, "value": "4.8 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.005, "value": "$0.005 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "act_agent", "label": "Act · NL-agent", "composite": 5.66, "rank": 2, "of": 2, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 0.667, "value": "0.67" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 70171.1, "value": "70.2 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.05, "value": "$0.05 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "scrapingbee": { "name": "ScrapingBee", "group": "Scraping & Crawling", "category": "Single-call scraping API", "tier": "T2", "route": "/scrapingbee", "call": "https://api.nativeport.ai/scrapingbee/", "docs": "https://docs.nativeport.ai/scrapingbee", "page": "https://nativeport.ai/providers/scrapingbee/", "pricing_entry": "~$0.98 / 1k (JS)", "latency_note": null, "provider_mcp": "official", "summary": "A single REST surface: JS rendering, residential and stealth proxies, extraction via AI or CSS, markdown out.", "choose_if": "For teams that want stealth, rendering, declarative actions and extraction consolidated into one endpoint.", "avoid_if": "The catch: costs climb quickly — JS rendering defaults to on at 5 credits, and stealth runs 75.", "evals": [ { "verb": "extract_rules", "label": "Extract · CSS rules", "composite": 9.23, "rank": 1, "of": 3, "top": true, "measured": "2026-06-22", "note": "Top-ranked rule-based extractor, at a fraction of AI-extract cost.", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2740.9, "value": "2.7 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.00098, "value": "$0.00098 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "act", "label": "Act · declarative", "composite": 8.72, "rank": 1, "of": 6, "top": true, "measured": "2026-06-22", "note": "Declarative browser actions: strongest showing.", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 5505.1, "value": "5.5 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.006, "value": "$0.006 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "screenshot", "label": "Screenshot", "composite": 8.57, "rank": 3, "of": 8, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 0.875, "value": "0.88" }, { "key": "full_page_support", "label": "Full-page", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2658.3, "value": "2.7 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.0028, "value": "$0.0028 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 12.5, "value": "12%" } ] }, { "verb": "scrape_domain", "label": "Scrape-domain", "composite": 8.19, "rank": 3, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "value_accuracy", "label": "Accuracy", "raw": 0.895, "value": "0.90" }, { "key": "field_fill", "label": "Field fill", "raw": 0.947, "value": "0.95" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 3591.4, "value": "3.6 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.01095, "value": "$0.011 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "extract_ai", "label": "Extract · AI/schema", "composite": 8.14, "rank": 3, "of": 4, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 0.893, "value": "0.89" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 3784.0, "value": "3.8 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.006719, "value": "$0.0067 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "scrape", "label": "Scrape", "composite": 7.45, "rank": 5, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.875, "value": "88%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 4.14, "value": "4.1 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 8929.9, "value": "8.9 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.005735, "value": "$0.0057 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "scrapfly": { "name": "Scrapfly", "group": "Scraping & Crawling", "category": "Single-call scraping API", "tier": "T1", "route": "/scrapfly", "call": "https://api.nativeport.ai/scrapfly/", "docs": "https://docs.nativeport.ai/scrapfly", "page": "https://nativeport.ai/providers/scrapfly/", "pricing_entry": "~$3.37–4.13 / 1k", "latency_note": null, "provider_mcp": "official", "summary": "Strong anti-bot bypass behind a single flag (asp=true), markdown and AI output, and a per-response cost breakdown.", "choose_if": "Pick it when anti-bot bypass that leads the benchmark, paired with the top measured AI-extraction accuracy, is the priority.", "avoid_if": "Not the pick if crawling or credit rollover matters; expect hard targets to cost around 30 credits per request.", "evals": [ { "verb": "scrape", "label": "Scrape", "composite": 9.48, "rank": 2, "of": 10, "top": false, "measured": "2026-06-20", "note": "Passes each anti-bot vendor with the single asp=true flag.", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 1.0, "value": "100%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 9.62, "value": "9.6 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4500.6, "value": "4.5 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.002475, "value": "$0.0025 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "extract_ai", "label": "Extract · AI/schema", "composite": 8.96, "rank": 1, "of": 4, "top": true, "measured": "2026-06-21", "note": "Extraction accuracy ranks first among the set.", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4508.4, "value": "4.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.002, "value": "$0.002 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "screenshot", "label": "Screenshot", "composite": 8.56, "rank": 4, "of": 8, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 1.0, "value": "1.00" }, { "key": "full_page_support", "label": "Full-page", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 11967.6, "value": "12.0 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.002, "value": "$0.002 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "act", "label": "Act · declarative", "composite": 7.01, "rank": 4, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 0.8, "value": "0.80" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 17445.6, "value": "17.4 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.002, "value": "$0.002 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "zenrows": { "name": "ZenRows", "group": "Scraping & Crawling", "category": "Scraper API + CDP browser", "tier": "T2", "route": "/zenrows", "call": "https://api.nativeport.ai/zenrows/", "docs": "https://docs.nativeport.ai/zenrows", "page": "https://nativeport.ai/providers/zenrows/", "pricing_entry": "~$1.40 / 1k (JS)", "latency_note": null, "provider_mcp": "official", "summary": "One key spans a scraper API together with a CDP scraping browser; heavy anti-bot sits behind a lone flag.", "choose_if": "Reach for it if a full scraping browser plus a scraper API under a single key, billed on success, fits your setup.", "avoid_if": "Skip it when your targets sit on its blocklist — in the corpus it produced heavy errors on *.toscrape.com.", "evals": [ { "verb": "scrape", "label": "Scrape", "composite": 7.31, "rank": 6, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.625, "value": "62%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 9.4, "value": "9.4 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4776.8, "value": "4.8 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.004704, "value": "$0.0047 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 25.0, "value": "25%" } ] }, { "verb": "act", "label": "Act · declarative", "composite": 6.89, "rank": 5, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 0.8, "value": "0.80" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 10132.8, "value": "10.1 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.005, "value": "$0.005 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "extract_rules", "label": "Extract · CSS rules", "composite": 5.51, "rank": 3, "of": 3, "top": false, "measured": "2026-06-22", "note": "Every *.toscrape.com domain is blocklisted — an openly stated gap.", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 0.5, "value": "0.50" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4492.7, "value": "4.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.0028, "value": "$0.0028 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 50.0, "value": "50%" } ] }, { "verb": "screenshot", "label": "Screenshot", "composite": 2.47, "rank": 8, "of": 8, "top": false, "measured": "2026-06-21", "note": "books.toscrape blocks it — a gap openly noted.", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 0.25, "value": "0.25" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 23258.2, "value": "23.3 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.0014, "value": "$0.0014 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 75.0, "value": "75%" } ] } ] }, "crawlbase": { "name": "Crawlbase", "group": "Scraping & Crawling", "category": "Single-call + managed crawl", "tier": "T2", "route": "/crawlbase", "call": "https://api.nativeport.ai/crawlbase/", "docs": "https://docs.nativeport.ai/crawlbase", "page": "https://nativeport.ai/providers/crawlbase/", "pricing_entry": "$3 / 1k std (success ×complexity)", "latency_note": null, "provider_mcp": "community", "summary": "Structured scrapers and markdown output alongside a managed asynchronous crawl queue; billing follows success, scaled by domain complexity.", "choose_if": "The right call when a managed crawl queue with success-based billing, priced by domain complexity, is what the job needs.", "avoid_if": "The catch: behavior is hard to predict — separate Normal and JS tokens, plus a body field whose type varies.", "evals": [ { "verb": "scrape_domain", "label": "Scrape-domain", "composite": 7.17, "rank": 5, "of": 6, "top": false, "measured": "2026-06-22", "note": "Instagram's anti-bot triggers non-deterministic behavior.", "metrics": [ { "key": "value_accuracy", "label": "Accuracy", "raw": 0.711, "value": "0.71" }, { "key": "field_fill", "label": "Field fill", "raw": 0.877, "value": "0.88" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 4684.1, "value": "4.7 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.002813, "value": "$0.0028 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "screenshot", "label": "Screenshot", "composite": 7.15, "rank": 6, "of": 8, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 0.625, "value": "0.62" }, { "key": "full_page_support", "label": "Full-page", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 3742.8, "value": "3.7 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.0015, "value": "$0.0015 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 25.0, "value": "25%" } ] }, { "verb": "scrape", "label": "Scrape", "composite": 7.08, "rank": 8, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.75, "value": "75%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 4.0, "value": "4.0 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 3543.2, "value": "3.5 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.006, "value": "$0.006 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 25.0, "value": "25%" } ] } ] }, "oxylabs": { "name": "Oxylabs", "group": "Scraping & Crawling", "category": "Scraping API + proxy network", "tier": "T2", "route": "/oxylabs", "call": "https://api.nativeport.ai/oxylabs/", "docs": "https://docs.nativeport.ai/oxylabs", "page": "https://nativeport.ai/providers/oxylabs/", "pricing_entry": "~$0.80–1.00 / 1k", "latency_note": null, "provider_mcp": "official", "summary": "Enterprise scraping backed by a proxy network in 195-plus countries, dedicated parsers, and success-only billing.", "choose_if": "Pick it for enterprise-grade scraping, parsers dedicated to Amazon and Google, or its NL-agent mode, which leads the eval.", "avoid_if": "Not the pick if crawl enumeration scoped to a path is needed — the AI-prompt crawler cannot deliver that.", "evals": [ { "verb": "scrape_domain", "label": "Scrape-domain", "composite": 8.74, "rank": 1, "of": 6, "top": true, "measured": "2026-06-22", "note": "Leading scraper, though official first-party APIs win where available.", "metrics": [ { "key": "value_accuracy", "label": "Accuracy", "raw": 0.93, "value": "0.93" }, { "key": "field_fill", "label": "Field fill", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2838.0, "value": "2.8 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.002151, "value": "$0.0022 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "extract_rules", "label": "Extract · CSS rules", "composite": 8.52, "rank": 2, "of": 3, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 9603.9, "value": "9.6 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.0012, "value": "$0.0012 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "act_agent", "label": "Act · NL-agent", "composite": 7.99, "rank": 1, "of": 2, "top": true, "measured": "2026-06-22", "note": "Front-runner in autonomous NL-agent mode.", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 47261.3, "value": "47.3 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.05, "value": "$0.05 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "scrape", "label": "Scrape", "composite": 7.1, "rank": 7, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.75, "value": "75%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 4.0, "value": "4.0 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 10534.3, "value": "10.5 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.001267, "value": "$0.0013 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "crawl", "label": "Crawl", "composite": 3.18, "rank": 5, "of": 5, "top": false, "measured": "2026-06-22", "note": "Path-scoped enumeration is beyond the AI-prompt crawler.", "metrics": [ { "key": "coverage", "label": "Coverage", "raw": 0.265, "value": "0.27" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 35526.2, "value": "35.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.127358, "value": "$0.13 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 50.0, "value": "50%" } ] } ] }, "brightdata": { "name": "Bright Data", "group": "Scraping & Crawling", "category": "Web Unlocker + Scraper API", "tier": "T1", "route": "/brightdata", "call": "https://api.nativeport.ai/brightdata/", "docs": "https://docs.nativeport.ai/brightdata", "page": "https://nativeport.ai/providers/brightdata/", "pricing_entry": "~$1.30–1.50 / 1k", "latency_note": null, "provider_mcp": "official", "summary": "Measured strongest against anti-bot — 98% success on the hardest sites — with pre-built structured scrapers and a top compliance posture.", "choose_if": "The right call when the toughest anti-bot targets are in scope and the strongest measured unblocker plus a serious compliance posture matter.", "avoid_if": "The catch: no crawl link-discovery — the url_collection dataset comes back with the seed URL alone.", "evals": [ { "verb": "scrape", "label": "Scrape", "composite": 7.82, "rank": 4, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.75, "value": "75%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 5.08, "value": "5.1 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 3852.7, "value": "3.9 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.001933, "value": "$0.0019 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "scrape_domain", "label": "Scrape-domain", "composite": 6.66, "rank": 6, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "value_accuracy", "label": "Accuracy", "raw": 0.722, "value": "0.72" }, { "key": "field_fill", "label": "Field fill", "raw": 0.75, "value": "0.75" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 8126.9, "value": "8.1 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.003463, "value": "$0.0035 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 25.0, "value": "25%" } ] }, { "verb": "crawl", "label": "Crawl", "composite": 4.47, "rank": 4, "of": 5, "top": false, "measured": "2026-06-22", "note": "No link discovery: url_collection hands back just the seed.", "metrics": [ { "key": "coverage", "label": "Coverage", "raw": 0.1, "value": "0.10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 6530.6, "value": "6.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.01, "value": "$0.01 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "spider": { "name": "Spider", "group": "Scraping & Crawling", "category": "Crawl / AI-focused platform", "tier": "T2", "route": "/spider", "call": "https://api.nativeport.ai/spider/", "docs": "https://docs.nativeport.ai/spider", "page": "https://nativeport.ai/providers/spider/", "pricing_entry": "~$0.03 / 1k (usage-based)", "latency_note": null, "provider_mcp": "official (22 tools)", "summary": "Recursive crawling at speed, markdown that comes back clean, and per-page USD costs reported inside the response.", "choose_if": "For teams that value crawl speed, clean markdown, and per-page pricing with no multipliers to decode.", "avoid_if": "Skip it when declarative actions matter — clicks no-op in live runs — or when usage-based billing must be forecastable.", "evals": [ { "verb": "crawl", "label": "Crawl", "composite": 8.54, "rank": 2, "of": 5, "top": false, "measured": "2026-06-22", "note": "Quick at recursive crawling.", "metrics": [ { "key": "coverage", "label": "Coverage", "raw": 0.771, "value": "0.77" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2888.4, "value": "2.9 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 6.2e-05, "value": "$6.2e-05 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "parse", "label": "Parse · PDF/doc", "composite": 6.99, "rank": 3, "of": 3, "top": false, "measured": "2026-06-22", "note": "Scanned PDFs yield raw bytes.", "metrics": [ { "key": "text_accuracy", "label": "Text accuracy", "raw": 0.6, "value": "0.60" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 2006.3, "value": "2.0 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 3.8e-05, "value": "$3.8e-05 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "scrape", "label": "Scrape", "composite": 6.73, "rank": 9, "of": 10, "top": false, "measured": "2026-06-20", "note": "", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 0.5, "value": "50%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 9.62, "value": "9.6 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 12120.0, "value": "12.1 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.00013, "value": "$0.00013 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "screenshot", "label": "Screenshot", "composite": 5.03, "rank": 7, "of": 8, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 0.5, "value": "0.50" }, { "key": "full_page_support", "label": "Full-page", "raw": 0.5, "value": "0.50" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 8410.0, "value": "8.4 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.001, "value": "$0.001 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 37.5, "value": "38%" } ] }, { "verb": "act", "label": "Act · declarative", "composite": 1.5, "rank": 6, "of": 6, "top": false, "measured": "2026-06-22", "note": "Live declarative clicks do nothing.", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 0.0, "value": "0.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 35091.3, "value": "35.1 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.001, "value": "$0.001 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "zyte": { "name": "Zyte", "group": "Scraping & Crawling", "category": "Single-endpoint extraction", "tier": "T1", "route": "/zyte", "call": "https://api.nativeport.ai/zyte/", "docs": "https://docs.nativeport.ai/zyte", "page": "https://nativeport.ai/providers/zyte/", "pricing_entry": "$0.13–$1.27 / 1k (HTTP)", "latency_note": null, "provider_mcp": "community", "summary": "A single endpoint pairing automatic ban-management with AI extraction typed for product, article and job pages.", "choose_if": "Pick it when ban-management should be automatic and product, article or job pages need typed AI extraction.", "avoid_if": "Not the pick if extraction must include custom fields; the schema is fixed and won't accommodate them.", "evals": [ { "verb": "screenshot", "label": "Screenshot", "composite": 9.04, "rank": 2, "of": 8, "top": false, "measured": "2026-06-21", "note": "", "metrics": [ { "key": "valid_image_rate", "label": "Valid image", "raw": 1.0, "value": "1.00" }, { "key": "full_page_support", "label": "Full-page", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 5874.5, "value": "5.9 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.001, "value": "$0.001 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "scrape", "label": "Scrape", "composite": 8.63, "rank": 3, "of": 10, "top": false, "measured": "2026-06-20", "note": "Gets past every anti-bot vendor.", "metrics": [ { "key": "block_bypass_success_rate", "label": "Anti-bot bypass", "raw": 1.0, "value": "100%" }, { "key": "markdown_cleanliness", "label": "Markdown clean", "raw": 4.25, "value": "4.2 / 10" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 6518.4, "value": "6.5 s" }, { "key": "cost_per_successful_page_usd", "label": "Cost", "raw": 0.000687, "value": "$0.00069 / success" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "act", "label": "Act · declarative", "composite": 8.5, "rank": 2, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "task_success", "label": "Task success", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 11846.7, "value": "11.8 s" }, { "key": "cost_per_call_usd", "label": "Cost", "raw": 0.001, "value": "$0.001 / call" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] }, { "verb": "extract_ai", "label": "Extract · AI/schema", "composite": 5.7, "rank": 4, "of": 4, "top": false, "measured": "2026-06-21", "note": "Custom fields aren't possible under its fixed schema.", "metrics": [ { "key": "field_accuracy", "label": "Field accuracy", "raw": 0.55, "value": "0.55" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 6503.6, "value": "6.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.001818, "value": "$0.0018 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "apify": { "name": "Apify", "group": "Browser Automation", "category": "Actor marketplace", "tier": "T2", "route": "/apify", "call": "https://api.nativeport.ai/apify/", "docs": "https://docs.nativeport.ai/apify", "page": "https://nativeport.ai/providers/apify/", "pricing_entry": "varies (CU + rental + result)", "latency_note": null, "provider_mcp": "official", "summary": "Scraping and automation via a 7,000+ Actor marketplace, consolidated on one key and one bill.", "choose_if": "Reach for it if an Actor already exists for your target site or task and building a scraper can be skipped.", "avoid_if": "The catch: there is no universal response schema, since every Actor defines its own item shape.", "evals": [ { "verb": "scrape_domain", "label": "Scrape-domain", "composite": 8.23, "rank": 2, "of": 6, "top": false, "measured": "2026-06-22", "note": "", "metrics": [ { "key": "value_accuracy", "label": "Accuracy", "raw": 0.95, "value": "0.95" }, { "key": "field_fill", "label": "Field fill", "raw": 1.0, "value": "1.00" }, { "key": "latency_p50_ms", "label": "Latency p50", "raw": 8533.8, "value": "8.5 s" }, { "key": "cost_per_useful_usd", "label": "Cost", "raw": 0.003158, "value": "$0.0032 / useful" }, { "key": "error_rate_pct", "label": "Errors", "raw": 0.0, "value": "0%" } ] } ] }, "browserbase": { "name": "Browserbase", "group": "Browser Automation", "category": "Browser-as-a-service", "tier": "T2", "route": "/browserbase", "call": "https://api.nativeport.ai/browserbase/", "docs": "https://docs.nativeport.ai/browserbase", "page": "https://nativeport.ai/providers/browserbase/", "pricing_entry": "$0.12 / browser-hr (+ GB)", "latency_note": null, "provider_mcp": "official (Stagehand)", "summary": "Cloud headless browser sessions, driven through Playwright or Puppeteer, suited to auth-walled and multi-step flows.", "choose_if": "The right call when flows are stateful, multi-step or behind a login, and driving the browser yourself is the plan.", "avoid_if": "Not the pick for a one-shot URL-to-content call: what comes back is a session handle rather than content.", "evals": [], "gated": [ { "verbs": [ "act", "act_agent", "screenshot" ], "label": "Act / Screenshot", "note": "Evaluated but unscored: Browserbase hands you a CDP browser session to drive, not a synchronous goal-in/content-out REST call, so the request/response corpora can't exercise it without a stateful adapter." } ] }, "browserless": { "name": "Browserless", "group": "Browser Automation", "category": "Stateless browser processing", "tier": null, "route": "/browserless", "call": "https://api.nativeport.ai/browserless/", "docs": "https://docs.nativeport.ai/browserless", "page": "https://nativeport.ai/providers/browserless/", "pricing_entry": "$0.004 / accepted forwarded attempt", "latency_note": "≤30s bound per call", "provider_mcp": "none", "summary": "One call in, a rendered result out: screenshot, PDF, HTML, or a CSS-selector scrape from a single page load.", "choose_if": "Reach for it when the job is a single page — grab a screenshot, PDF, rendered HTML, or a structured CSS-selector scrape — and there's no session to keep alive.", "avoid_if": "Not the pick for a multi-step, stateful flow: driving a Playwright/Puppeteer session yourself, logins, or anything spanning more than one page load sits outside this route.", "evals": [], "gated": [ { "verbs": [ "screenshot", "extract_rules" ], "label": "Screenshot / Extract · CSS rules", "note": "Not yet benchmarked: the newly added Browserless fixed REST surface has not entered the current benchmark snapshot, so no score is published yet." } ] }, "steel": { "name": "Steel", "group": "Browser Automation", "category": "Stateless browser processing", "tier": null, "route": "/steel", "call": "https://api.nativeport.ai/steel/", "docs": "https://docs.nativeport.ai/steel", "page": "https://nativeport.ai/providers/steel/", "pricing_entry": "$0.005 / forwarded call", "latency_note": null, "provider_mcp": "official (steel-mcp-server, drives Steel's own CDP sessions — not this route)", "summary": "Three fixed actions — scrape to Markdown/HTML, screenshot, or PDF — one page load in, one response out.", "choose_if": "Reach for it when the job is one URL and the output is page content, a screenshot, or a PDF — no session to keep alive, and no query-string configuration to get right.", "avoid_if": "Not the pick for a stateful flow: logins, multi-step interaction, a live session to drive, or Steel's own proxy/CAPTCHA handling all sit outside this route.", "evals": [], "gated": [ { "verbs": [ "scrape", "screenshot" ], "label": "Scrape / Screenshot", "note": "Not yet benchmarked: the newly added Steel Browser Tools surface has not entered the current benchmark snapshot, so no score is published yet." } ] }, "elevenlabs": { "name": "ElevenLabs", "group": "Voice & Audio", "category": "Text-to-speech & voice", "tier": null, "route": "/elevenlabs", "call": "https://api.nativeport.ai/elevenlabs/", "docs": "https://docs.nativeport.ai/elevenlabs", "page": "https://nativeport.ai/providers/elevenlabs/", "pricing_entry": "see ElevenLabs pricing", "latency_note": null, "provider_mcp": "official", "summary": "Wildcard routing exposes the complete ElevenLabs REST surface for voice and text-to-speech.", "choose_if": "Pick it when voice cloning or high-quality text-to-speech should run on the key you already use.", "avoid_if": null, "evals": [] }, "fishaudio": { "name": "Fish Audio", "group": "Voice & Audio", "category": "TTS, transcription & voice design", "tier": null, "route": "/fishaudio", "call": "https://api.nativeport.ai/fishaudio/", "docs": "https://docs.nativeport.ai/fishaudio", "page": "https://nativeport.ai/providers/fishaudio/", "pricing_entry": "$15/1M UTF-8 bytes TTS / $0.36/audio-hr ASR / $0.01/req voice-design", "latency_note": null, "provider_mcp": "community", "summary": "Four fixed routes reach Fish Audio's text-to-speech, transcription, and from-a-description voice-design models, each metered on the exact unit Fish Audio itself bills — bytes, audio-hours, or a flat per-call rate.", "choose_if": "Pick it for production text-to-speech with expressive, natural-sounding output, real speech-to-text, or building a new voice from a written description, all landing on the same key and balance as everything else here.", "avoid_if": "Skip it for persistent voice-clone models or account-balance visibility: Fish Audio's model-management and wallet endpoints aren't mounted, since a shared gateway key would let one customer see or reuse another's saved models and balance. Per-request zero-shot cloning still works on a single `/v1/tts` call, and Fish Audio's WebSocket streaming endpoint isn't proxied either — its chunked, timestamped HTTP streaming route covers the same low-latency case instead.", "evals": [] }, "openai": { "name": "OpenAI", "group": "Model Inference", "category": "Chat, embeddings & moderation API", "tier": null, "route": "/openai", "call": "https://api.nativeport.ai/openai/", "docs": "https://docs.nativeport.ai/openai", "page": "https://nativeport.ai/providers/openai/", "pricing_entry": "$0.02 → $45 / 1M tokens (gpt-5.6 family: Sol/Terra/Luna, cache- & context-tiered; embeddings/moderation cheaper or free)", "latency_note": null, "provider_mcp": "none", "summary": "OpenAI's Responses, Chat Completions, embeddings and moderation endpoints, metered per token against the same balance as the rest of the catalog, with gpt-5.6's cache- and context-tiered Standard rates read straight from each response.", "choose_if": "Pick it when the task calls for an OpenAI model by name and per-token spend should land on the one ledger you already watch.", "avoid_if": "Skip it for anything outside the gateway's admitted surface: only a fixed, maintained model list and a closed set of request fields are reachable, and everything else is rejected before it reaches OpenAI.", "evals": [] }, "anthropic": { "name": "Anthropic", "group": "Model Inference", "category": "Claude Messages API", "tier": null, "route": "/anthropic", "call": "https://api.nativeport.ai/anthropic/", "docs": "https://docs.nativeport.ai/anthropic", "page": "https://nativeport.ai/providers/anthropic/", "pricing_entry": "$1 → $25 / 1M tokens (per model, cache-tiered)", "latency_note": null, "provider_mcp": "none", "summary": "Claude's Messages API behind the same policed-route pattern as the OpenAI proxy, cache-tier pricing included.", "choose_if": "Pick it for Claude's models specifically, with prompt-cache pricing tiers read straight from each response and billed to the one balance.", "avoid_if": "Skip it for anything outside the gateway's admitted surface: only a fixed, current Claude model list and a closed set of request fields are admitted here.", "evals": [] }, "grok": { "name": "Grok", "group": "Model Inference", "category": "xAI's Grok chat & Responses API", "tier": null, "route": "/grok", "call": "https://api.nativeport.ai/grok/", "docs": "https://docs.nativeport.ai/grok", "page": "https://nativeport.ai/providers/grok/", "pricing_entry": "$1 → $12 / 1M tokens (per model, context-tiered; dollar-direct metering)", "latency_note": null, "provider_mcp": "none", "summary": "xAI's Grok chat and Responses API, billed from the dollar cost each response reports rather than a maintained price table.", "choose_if": "Pick it for Grok's models specifically, where the real per-call cost rides in the response itself instead of a separately kept price list.", "avoid_if": "Skip it for anything outside the gateway's admitted surface: only a fixed model list and a closed set of request fields are reachable, same as the other model-inference routes here.", "evals": [] }, "huggingface": { "name": "Hugging Face", "group": "Model Inference", "category": "Multi-provider chat inference router", "tier": null, "route": "/huggingface", "call": "https://api.nativeport.ai/huggingface/", "docs": "https://docs.nativeport.ai/huggingface", "page": "https://nativeport.ai/providers/huggingface/", "pricing_entry": "$0.05 → $4.40 / 1M tokens (per model+provider pin, pass-through)", "latency_note": null, "provider_mcp": "none", "summary": "Hugging Face's Inference Providers router, fanned out across competing serving providers behind one chat-completions endpoint.", "choose_if": "Pick it to reach open models served by whichever backing provider is cataloged for them, without separate accounts or price lists per provider.", "avoid_if": "The leanest of the model-inference routes: no files, batches or fine-tuning surface, and no access to the router's own fastest/cheapest/preferred routing — chat completions only, pinned to specific cataloged (model, provider) pairs.", "evals": [] } } }