---
title: "Best company funding data API (2026): startup funding providers ranked by an independent benchmark | Openbenchmarks"
description: "Which company funding data provider is best? An independent 2026 benchmark scoring 17 providers on whether they return the correct latest funding stage, against funding events verified from official sources. Two boards: freshness for rounds announced in the last 30 days, enrichment for older rounds. Every figure comes from a run you can reproduce. Open data and code."
url: https://openbenchmarks.com/company-funding
---

[](https://openbenchmarks.com/) 

[← all benchmarks](https://openbenchmarks.com/) 

[benchmarks](https://openbenchmarks.com/) / company funding 

04 · company funding

# Company Funding Benchmark

**What a funding data provider is asked for.** Given a company domain, return that company's **latest funding stage**: the most recent round it raised, named correctly.

**Two boards.** [Freshness](#results) measures rounds announced in the last 30 days: how fast a provider indexes news. Enrichment measures rounds older than that: how completely it has backfilled history. These reward opposite things, and a provider that leads one routinely trails the other, so they are ranked separately rather than averaged into a rank that describes neither. Each freshness cohort ages into enrichment once its window closes.

**Vendors benchmarked.** The 17 measured providers are Apollo, CompanyEnrich, Crunchbase, Crustdata, Exa, Explorium, Fiber, Firecrawl, Fundable, Harmonic, Nimble, Ocean.io, Parallel, People Data Labs, PredictLeads, Seltz, and ZoomInfo. Several are measured on more than one endpoint, for 23 arms in total. Crunchbase and Harmonic are measured from reviewed exports mapped to the same company identities, so neither carries a latency or cost figure. PitchBook and Tracxn are private-market research platforms built for investor workflows and hence are not benched here.

**Which board answers your question.** Start from what consumes the data instead of the top of a ranking. Vendors are divided in 3 categories based on how they source data: **Long Running Agent APIs** research each company at request time, **Web search APIs** extract it from a search index, and **GTM data providers** return a stored record.

* **Freshness of Funding Data.** Only accuracy matters in freshness on recent rounds. All three kinds of providers have comparable scores, so pick on price.
* **CRM Backfill Enrichment.** Speed does not matter in a backfill enrichment job, so measure on accuracy and price. Read the enrichment board along with the funding fields returned, where the long running agents are the most accurate.
* **Enrichment at Scale.** Price matters here the most. Compare correct stages per dollar across the web search APIs and GTM data providers. Long running agents are the most expensive and are not recommended for running enrichment at scale.
* **Live Agent Enrichment Workflow.** Speed and accuracy matter the most here. Read median latency first, then correct when returned, and only among web search APIs and GTM providers. Long running agent APIs are significantly slower than the other two types of providers, and are not recommended for this usecase.

2026-09-15 added Nimble Search lite and standard to the enrichment and freshness boards, last run 2026-09-15. [Full changelog →](#changelog) 

[open data + code Inputs, normalized provider outputs, and evaluation code are public in openbenchmarks-labs/company-funding. github → ](https://github.com/openbenchmarks-labs/company-funding "Open the public data + code mirror on GitHub") 

[Firecrawlagentic web extraction, Spark 1 Mini](https://firecrawl.dev "Open Firecrawl in a new tab") 

[Parallelagentic research, Task API](https://parallel.ai "Open Parallel in a new tab") 

[Parallelagentic research, Responses API](https://parallel.ai "Open Parallel in a new tab") 

[Exaagentic research, Agent API](https://exa.ai "Open Exa in a new tab") 

[Exaweb embeddings, deep reasoning](https://exa.ai "Open Exa in a new tab") 

[Firecrawlagentic web extraction, Spark 2](https://firecrawl.dev "Open Firecrawl in a new tab") 

[Exaweb embeddings, instant search](https://exa.ai "Open Exa in a new tab") 

[Crunchbasefunding database](https://www.crunchbase.com "Open Crunchbase in a new tab") 

[Fibercompany search API](https://www.fiber.ai "Open Fiber in a new tab") 

[NimbleSearch API, standard](https://www.nimbleway.com "Open Nimble in a new tab") 

[Crustdatacompany data API](https://crustdata.com "Open Crustdata in a new tab") 

[Harmonicstartup data platform](https://www.harmonic.ai "Open Harmonic in a new tab") 

[People Data Labscompany dataset](https://www.peopledatalabs.com "Open People Data Labs in a new tab") 

[Fundablefunding data API](https://www.tryfundable.ai "Open Fundable in a new tab") 

[NimbleSearch API, lite](https://www.nimbleway.com "Open Nimble in a new tab") 

[Seltzanswer API, companies scope](https://seltz.ai "Open Seltz in a new tab") 

[ApolloB2B sales database](https://www.apollo.io "Open Apollo in a new tab") 

[CompanyEnrichcompany enrichment API](https://companyenrich.com "Open CompanyEnrich in a new tab") 

[ZoomInfoGTM data platform](https://gtm.ai "Open ZoomInfo in a new tab") 

[Seltzanswer API, news scope](https://seltz.ai "Open Seltz in a new tab") 

[PredictLeadsco-signal graph](https://predictleads.com "Open PredictLeads in a new tab") 

[Exploriumenrichment platform](https://www.explorium.ai "Open Explorium in a new tab") 

[Ocean.iocompany graph](https://www.ocean.io "Open Ocean.io in a new tab") 

\[01\] results

## Compare latest-stage accuracy, returned data, and speed.

Every column is a measured value. Column definitions, the matching policy, and what each number does not tell you are in [\[02\] methodology](#methodology).

### Funding freshness: rounds announced in the last 30 days

Measured on 181 company observations across 4 dated snapshots: 2026-09 (2026-08-14 to 2026-09-13), 2026-08 (2026-07-27 to 2026-08-26), 2026-08 (2026-07-16 to 2026-08-15) and 2026-07 (2026-06-16 to 2026-07-15). This is a rolling measurement: every vendor is re-run on a new snapshot of recent rounds, so these numbers change over time. Each snapshot above is scored on the rounds announced inside its own window. 

benchmarks/funding/freshness rolling 

Top 8 per metric 

__Funding providers compared on rounds announced in the last 30 days (window ending 2026-09-13)__
| Provider | Endpoint & configuration | Measured on | Latest stage correct | Correct when returned | Latest stage returned | Funding fields returned | Fields returned | Median latency | Est. cost | Official docs | Company identified |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Long Running Agent APIs | | | | | | | | | | | |
| [ Exa ](https://exa.ai) | POST /agent/runs effort=medium · JSON schema | **100** company measurements · 2 snapshots | 97.0% | 97.9% | 96.0% | 73.6% | 3.7 / 5 | 27,293 ms | **$10.00** $0.10 / request (medium effort) | [Exa official docs ↗ ](https://exa.ai/docs/reference/agent-api-guide) | 100.0% 100/100 domains |
| [ Firecrawl ](https://firecrawl.dev) | POST /v2/agent model=spark-2 · JSON schema · maxCredits capped per run | **50** company measurements · 1 snapshot | 100.0% | 100.0% | 98.0% | 87.2% | 4.4 / 5 | 102,203 ms | **$1.98** $0.59 / 1,000 credits | [Firecrawl official docs ↗ ](https://docs.firecrawl.dev/features/agent) | 100.0% 50/50 domains |
| [ Firecrawl ](https://firecrawl.dev) | POST /v2/agent model=spark-1-mini · JSON schema · maxCredits capped per run | **100** company measurements · 2 snapshots | 96.0% | 96.9% | 97.0% | 94.4% | 4.7 / 5 | 92,998 ms | **$9.95** $0.59 / 1,000 credits | [Firecrawl official docs ↗ ](https://docs.firecrawl.dev/features/agent) | 100.0% 100/100 domains |
| [ Parallel ](https://parallel.ai) | POST /v1/tasks/runs processor=core · structured JSON output | **181** company measurements · 3 snapshots | 95.0% | 96.0% | 95.6% | 86.7% | 4.3 / 5 | 62,142 ms | **$4.53** $25 / 1,000 Task runs | [Parallel official docs ↗ ](https://docs.parallel.ai/api-reference/tasks/create-task-run) | 100.0% 181/181 domains |
| Web search APIs | | | | | | | | | | | |
| [ Exa ](https://exa.ai) | POST /search type=instant · JSON schema | **100** company measurements · 2 snapshots | 98.0% | 99.0% | 95.0% | 90.2% | 4.5 / 5 | 1,618 ms | **$0.70** $0.007 / request | [Exa official docs ↗ ](https://exa.ai/docs/reference/search) | 100.0% 100/100 domains |
| [ Exa ](https://exa.ai) | POST /search type=deep-reasoning · JSON schema | **181** company measurements · 3 snapshots | 95.6% | 96.0% | 96.7% | 84.8% | 4.2 / 5 | 9,916 ms | **$2.71** $0.015 / request | [Exa official docs ↗ ](https://exa.ai/docs/reference/search) | 100.0% 181/181 domains |
| [ Nimble ](https://www.nimbleway.com) | POST /v2/search search\_depth=standard · focus=general · full\_content=false · 10 results · source-only funding extraction | **46** company measurements · 1 snapshot | 84.8% | 94.6% | 80.4% | 67.8% | 3.4 / 5 | 3,907 ms | **$1.18** $5 / 1,000 searches + extraction tokens | [Nimble official docs ↗ ](https://docs.nimbleway.com/api-reference/search/search) | 95.7% 44/46 domains |
| [ Nimble ](https://www.nimbleway.com) | POST /v2/search search\_depth=lite · focus=general · full\_content=false · 10 results · source-only funding extraction | **46** company measurements · 1 snapshot | 63.0% | 92.9% | 60.9% | 38.7% | 1.9 / 5 | 3,829 ms | **$0.23** $1.10 / 1,000 searches + extraction tokens | [Nimble official docs ↗ ](https://docs.nimbleway.com/api-reference/search/search) | 76.1% 35/46 domains |
| [ Parallel ](https://parallel.ai) | POST /v1/responses model=parallel · reasoning.effort=medium · text.format JSON schema | **100** company measurements · 2 snapshots | 90.0% | 90.6% | 96.0% | 82.6% | 4.1 / 5 | 17,020 ms | **$5.00** $50 / 1,000 Responses requests (medium reasoning) | [Parallel official docs ↗ ](https://docs.parallel.ai/api-reference/responses-api/create-response) | 100.0% 100/100 domains |
| [ Seltz ](https://seltz.ai) | POST /v1/answer scope=news · response\_format JSON | **100** company measurements · 2 snapshots | 71.0% | 85.9% | 78.0% | 57.8% | 2.9 / 5 | 2,109 ms | **\--** Pricing not disclosed | [Seltz official docs ↗ ](https://docs.seltz.ai/api-reference/answer) | 100.0% 100/100 domains |
| [ Seltz ](https://seltz.ai) | POST /v1/answer scope=companies · response\_format JSON | **100** company measurements · 2 snapshots | 22.0% | 32.8% | 67.0% | 62.6% | 3.1 / 5 | 2,118 ms | **\--** Pricing not disclosed | [Seltz official docs ↗ ](https://docs.seltz.ai/api-reference/answer) | 100.0% 100/100 domains |
| GTM data providers | | | | | | | | | | | |
| [ Apollo ](https://www.apollo.io) | GET /api/v1/organizations/enrich | **181** company measurements · 3 snapshots | 59.7% | 69.5% | 85.1% | 70.1% | 3.5 / 5 | 320 ms | **$3.44** $0.0196 / modeled credit | [Apollo official docs ↗ ](https://docs.apollo.io/reference/organization-enrichment) | 97.2% 176/181 domains |
| [ CompanyEnrich ](https://companyenrich.com) | GET /companies/enrich | **181** company measurements · 3 snapshots | 12.7% | 19.1% | 63.5% | 56.8% | 2.8 / 5 | 334 ms | **$1.77** $0.0098 / credit | [CompanyEnrich official docs ↗ ](https://docs.companyenrich.com/reference/get%5Fcompanies-enrich) | 96.7% 175/181 domains |
| [ Crunchbase (exported dataset) ](https://www.crunchbase.com) | Crunchbase self-serve plan export identity-mapped export · not an API run | **81** company measurements · 1 snapshot | 95.1% | 98.7% | 96.3% | 92.6% | 4.6 / 5 | \-- | **\--** Not comparable | [Crunchbase (exported dataset) official docs ↗ ](https://data.crunchbase.com/docs/using-the-api) | 100.0% 81/81 domains |
| [ Crustdata ](https://crustdata.com) | POST /company/enrich exact\_match=true · fields=\[funding\] · 25 domains/batch | **181** company measurements · 3 snapshots | 81.2% | 90.1% | 89.5% | 71.3% | 3.6 / 5 | 1,537 ms | **$54.30** $0.30 / response credit | [Crustdata official docs ↗ ](https://docs.crustdata.com/company-docs/enrichment/introduction) | 96.1% 174/181 domains |
| [ Explorium ](https://www.explorium.ai) | POST /v1/businesses/match + bulk\_enrich funding\_and\_acquisition enrichment | **181** company measurements · 3 snapshots | 5.5% | 13.6% | 36.5% | 40.4% | 2.0 / 5 | 715 ms | **$11.00** $0.04 / request proxy | [Explorium official docs ↗ ](https://explorium-api.readme.io/reference/match%5Fbusinesses) | 44.8% 81/181 domains |
| [ Fiber ](https://www.fiber.ai) | POST /v1/company-search | **181** company measurements · 3 snapshots | 74.6% | 80.2% | 92.3% | 88.3% | 4.4 / 5 | 615 ms | **$4.11** $0.020 / credit | [Fiber official docs ↗ ](https://api.fiber.ai/docs/) | 97.8% 177/181 domains |
| [ Fundable ](https://www.tryfundable.ai) | GET /api/v1/company domain query parameter | **100** company measurements · 2 snapshots | 84.0% | 92.3% | 91.0% | 91.0% | 4.5 / 5 | 1,300 ms | **$3.53** $0.05 / credit | [Fundable official docs ↗ ](https://docs.tryfundable.ai/) | 91.0% 91/100 domains |
| [ Harmonic (exported dataset) ](https://www.harmonic.ai) | Supplied Harmonic company export identity-audited mapping · not an API run | **81** company measurements · 1 snapshot | 81.5% | 81.5% | 100.0% | 91.8% | 4.6 / 5 | \-- | **\--** Pricing not disclosed | [Harmonic (exported dataset) official docs ↗ ](https://console.harmonic.ai/docs/api-reference/introduction) | 100.0% 81/81 domains |
| [ Ocean.io ](https://www.ocean.io) | POST /v2/enrich/company | **181** company measurements · 3 snapshots | 11.1% | 19.4% | 54.1% | 29.4% | 1.5 / 5 | 575 ms | **$1.11** $0.064 / credit | [Ocean.io official docs ↗ ](https://docs.ocean.io) | 96.7% 175/181 domains |
| [ People Data Labs ](https://www.peopledatalabs.com) | GET /v5/company/enrich | **181** company measurements · 3 snapshots | 13.3% | 19.5% | 65.2% | 51.4% | 2.6 / 5 | 325 ms | **$17.38** $0.10 / successful match | [People Data Labs official docs ↗ ](https://docs.peopledatalabs.com/docs/reference-company-enrichment-api) | 92.3% 167/181 domains |
| [ PredictLeads ](https://predictleads.com) | GET /api/v3/companies/{domain}/financing\_events | **181** company measurements · 3 snapshots | 68.0% | 82.4% | 81.8% | 64.3% | 3.2 / 5 | 561 ms | **$4.83** $0.04 / request · 100 free | [PredictLeads official docs ↗ ](https://docs.predictleads.com/guide/extended%5Fcompanies%5Fdataset) | 84.0% 152/181 domains |
| [ ZoomInfo ](https://gtm.ai) | gtm companies enrich --file 10 domains/batch · 4 funding fields | **181** company measurements · 3 snapshots | 72.4% | 85.0% | 84.5% | 85.8% | 4.3 / 5 | 958 ms | **$18.10** $0.10 / credit launch promo · $0.35 standard | [ZoomInfo official docs ↗ ](https://gtm.ai/docs/mcp/tools/enrich-companies) | 90.6% 164/181 domains |

### Funding enrichment: rounds announced more than 30 days ago

The historical board. A vendor is scored on the rounds that were already more than 30 days old when it ran. Each row's own denominator is shown in the company identified column.

benchmarks/funding/company-funding-enrichment-v3 live 

Top 8 per metric 

__Funding information providers compared on latest funding-stage data__
| Provider | Endpoint & configuration | Latest stage correct | Correct when returned | Latest stage returned | Funding fields returned | Fields returned | Median latency | Est. cost | Official docs | Company identified |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Long Running Agent APIs | | | | | | | | | | |
| [ Exa ](https://exa.ai) | POST /agent/runs effort=medium · JSON schema | 88.7% | 93.0% | 90.7% | 67.3% | 3.4 / 5 | 26,549 ms | **$30.00** $0.10 / request (medium effort) | [Exa official docs ↗ ](https://exa.ai/docs/reference/agent-api-guide) | 100.0% 300/300 domains |
| [ Firecrawl ](https://firecrawl.dev) | POST /v2/agent model=spark-1-mini · JSON schema · maxCredits capped per run | 92.3% | 92.3% | 100.0% | 93.3% | 4.7 / 5 | 109,835 ms | **$29.86** $0.59 / 1,000 credits | [Firecrawl official docs ↗ ](https://docs.firecrawl.dev/features/agent) | 100.0% 300/300 domains |
| [ Firecrawl ](https://firecrawl.dev) | POST /v2/agent model=spark-2 · JSON schema · maxCredits capped per run | 87.3% | 89.9% | 95.3% | 84.2% | 4.2 / 5 | 104,406 ms | **$11.87** $0.59 / 1,000 credits | [Firecrawl official docs ↗ ](https://docs.firecrawl.dev/features/agent) | 99.7% 299/300 domains |
| [ Parallel ](https://parallel.ai) | POST /v1/tasks/runs processor=core · structured JSON output | 90.0% | 92.3% | 95.4% | 84.9% | 4.3 / 5 | 53,684 ms | **$5.47** $25 / 1,000 Task runs | [Parallel official docs ↗ ](https://docs.parallel.ai/api-reference/tasks/create-task-run) | 100.0% 219/219 domains |
| Web search APIs | | | | | | | | | | |
| [ Exa ](https://exa.ai) | POST /search type=deep-reasoning · JSON schema | 88.6% | 88.6% | 100.0% | 93.6% | 4.7 / 5 | 12,848 ms | **$3.29** $0.015 / request | [Exa official docs ↗ ](https://exa.ai/docs/reference/search) | 100.0% 219/219 domains |
| [ Exa ](https://exa.ai) | POST /search type=instant · JSON schema | 87.0% | 89.0% | 97.0% | 89.0% | 4.5 / 5 | 1,507 ms | **$2.10** $0.007 / request | [Exa official docs ↗ ](https://exa.ai/docs/reference/search) | 100.0% 300/300 domains |
| [ Nimble ](https://www.nimbleway.com) | POST /v2/search search\_depth=standard · focus=general · full\_content=false · 10 results · source-only funding extraction | 81.3% | 86.3% | 90.0% | 66.0% | 3.3 / 5 | 3,973 ms | **$7.82** $5 / 1,000 searches + extraction tokens | [Nimble official docs ↗ ](https://docs.nimbleway.com/api-reference/search/search) | 95.0% 285/300 domains |
| [ Nimble ](https://www.nimbleway.com) | POST /v2/search search\_depth=lite · focus=general · full\_content=false · 10 results · source-only funding extraction | 59.0% | 91.1% | 60.0% | 40.0% | 2.0 / 5 | 3,807 ms | **$1.51** $1.10 / 1,000 searches + extraction tokens | [Nimble official docs ↗ ](https://docs.nimbleway.com/api-reference/search/search) | 70.7% 212/300 domains |
| [ Parallel ](https://parallel.ai) | POST /v1/responses model=parallel · reasoning.effort=medium · text.format JSON schema | 89.0% | 90.1% | 94.3% | 81.6% | 4.1 / 5 | 15,930 ms | **$15.00** $50 / 1,000 Responses requests (medium reasoning) | [Parallel official docs ↗ ](https://docs.parallel.ai/api-reference/responses-api/create-response) | 100.0% 300/300 domains |
| [ Seltz ](https://seltz.ai) | POST /v1/answer scope=companies · response\_format JSON | 54.7% | 59.9% | 91.3% | 80.7% | 4.0 / 5 | 1,969 ms | **\--** Pricing not disclosed | [Seltz official docs ↗ ](https://docs.seltz.ai/api-reference/answer) | 100.0% 300/300 domains |
| [ Seltz ](https://seltz.ai) | POST /v1/answer scope=news · response\_format JSON | 40.0% | 59.3% | 66.3% | 46.3% | 2.3 / 5 | 1,926 ms | **\--** Pricing not disclosed | [Seltz official docs ↗ ](https://docs.seltz.ai/api-reference/answer) | 100.0% 300/300 domains |
| GTM data providers | | | | | | | | | | |
| [ Apollo ](https://www.apollo.io) | GET /api/v1/organizations/enrich | 49.8% | 63.7% | 78.1% | 64.3% | 3.2 / 5 | 284 ms | **$4.16** $0.0196 / modeled credit | [Apollo official docs ↗ ](https://docs.apollo.io/reference/organization-enrichment) | 96.8% 212/219 domains |
| [ CompanyEnrich ](https://companyenrich.com) | GET /companies/enrich | 46.1% | 55.2% | 83.6% | 69.3% | 3.5 / 5 | 342 ms | **$2.15** $0.0098 / credit | [CompanyEnrich official docs ↗ ](https://docs.companyenrich.com/reference/get%5Fcompanies-enrich) | 97.7% 214/219 domains |
| [ Crunchbase (exported dataset) ](https://www.crunchbase.com) | Crunchbase self-serve plan export identity-mapped export · not an API run | 85.8% | 90.8% | 94.5% | 87.8% | 4.4 / 5 | \-- | **\--** Not comparable | [Crunchbase (exported dataset) official docs ↗ ](https://data.crunchbase.com/docs/using-the-api) | 98.6% 216/219 domains |
| [ Crustdata ](https://crustdata.com) | POST /company/enrich exact\_match=true · fields=\[funding\] · 25 domains/batch | 79.0% | 89.2% | 88.6% | 68.4% | 3.4 / 5 | 1,469 ms | **$65.70** $0.30 / response credit | [Crustdata official docs ↗ ](https://docs.crustdata.com/company-docs/enrichment/introduction) | 94.1% 206/219 domains |
| [ Explorium ](https://www.explorium.ai) | POST /v1/businesses/match + bulk\_enrich funding\_and\_acquisition enrichment | 21.9% | 43.8% | 40.6% | 46.4% | 2.3 / 5 | 697 ms | **$13.32** $0.04 / request proxy | [Explorium official docs ↗ ](https://explorium-api.readme.io/reference/match%5Fbusinesses) | 53.4% 117/219 domains |
| [ Fiber ](https://www.fiber.ai) | POST /v1/company-search | 84.9% | 88.2% | 96.3% | 90.0% | 4.5 / 5 | 632 ms | **$4.98** $0.020 / credit | [Fiber official docs ↗ ](https://api.fiber.ai/docs/) | 97.3% 213/219 domains |
| [ Fundable ](https://www.tryfundable.ai) | GET /api/v1/company domain query parameter | 60.3% | 86.1% | 69.7% | 68.7% | 3.4 / 5 | 1,370 ms | **$10.60** $0.05 / credit | [Fundable official docs ↗ ](https://docs.tryfundable.ai/) | 70.7% 212/300 domains |
| [ Harmonic (exported dataset) ](https://www.harmonic.ai) | Supplied Harmonic company export identity-audited mapping · not an API run | 71.2% | 72.9% | 97.7% | 90.0% | 4.5 / 5 | \-- | **\--** Pricing not disclosed | [Harmonic (exported dataset) official docs ↗ ](https://console.harmonic.ai/docs/api-reference/introduction) | 98.6% 216/219 domains |
| [ Ocean.io ](https://www.ocean.io) | POST /v2/enrich/company | 5.5% | 14.3% | 35.2% | 21.1% | 1.1 / 5 | 551 ms | **$1.35** $0.064 / credit | [Ocean.io official docs ↗ ](https://docs.ocean.io) | 96.3% 211/219 domains |
| [ People Data Labs ](https://www.peopledatalabs.com) | GET /v5/company/enrich | 67.6% | 80.9% | 83.6% | 64.7% | 3.2 / 5 | 276 ms | **$21.02** $0.10 / successful match | [People Data Labs official docs ↗ ](https://docs.peopledatalabs.com/docs/reference-company-enrichment-api) | 95.9% 210/219 domains |
| [ PredictLeads ](https://predictleads.com) | GET /api/v3/companies/{domain}/financing\_events | 37.9% | 70.1% | 53.4% | 39.0% | 1.9 / 5 | 562 ms | **$5.84** $0.04 / request · 100 free | [PredictLeads official docs ↗ ](https://docs.predictleads.com/guide/extended%5Fcompanies%5Fdataset) | 57.5% 126/219 domains |
| [ ZoomInfo ](https://gtm.ai) | gtm companies enrich --file 10 domains/batch · 4 funding fields | 46.1% | 78.9% | 58.5% | 62.1% | 3.1 / 5 | 972 ms | **$21.90** $0.10 / credit launch promo · $0.35 standard | [ZoomInfo official docs ↗ ](https://gtm.ai/docs/mcp/tools/enrich-companies) | 76.7% 168/219 domains |

\[02\] methodology+ 

\[02.a\] cohort design

### Two boards, split on how old the round is

Finding a round announced this week and holding a correct historical record are different capabilities, and a vendor can be strong at one and weak at the other. Averaging them into a single number hides that, so they are measured separately. The split is on the announcement date: rounds inside the trailing 30 days are freshness, everything older is enrichment.

### Enrichment

Rounds announced more than 30 days ago. A durable historical set that grows each cycle as freshness snapshots age into it.

### Freshness

Rounds announced in the trailing 30 days, sourced from current public funding news and verified before inclusion. Measured while the rounds are still recent, which is what exposes update lag.

### Rolling windows

Each cycle builds a new dated freshness cohort and re-runs every vendor against it. Freshness results pool across the snapshots a vendor took part in, so its own case count is shown beside its score.

### Ground Truth reviewed

Latest-stage Ground Truth is reviewed per company against a primary source. Blank Ground Truth is a pass-through case.

\[02.b\] sourcing and validation

### Source first, provider second

1. Sample candidate companies from a time-windowed funding-transition index and independently research the announced event.
2. Keep records with a direct company newsroom, company-issued wire announcement, or equivalent primary source; retain source URL, date, stage, amount, and evidence note.
3. Build each freshness cohort from current public funding news inside that window, then require a verified company-domain identity before inclusion. A round with no verifiable announcement date is not placed in a window at all.
4. Freeze the input list before any provider is scored. Candidate-source fields are never used as provider outputs.

\[02.c\] evaluated field

### Latest funding stage

* Evaluate provider stages against the reviewed Ground Truth. The rubric handles documented equivalents, including Series suffixes within the same letter, PE/growth variants, strategic variants, and Seed Bridge/Pre-Series A.
* Series letters remain strict: a Series B variant can match Series B, but Seed and Pre-Seed remain distinct. Vague labels such as “early stage” do not match a specific Ground Truth stage.
* For non-Series stage disagreements, an exactly matching announced date or round amount can establish correctness.

\[02.d\] headline metric

### Latest stage correct

* `correct latest stages ÷ the Ground Truth-reviewed companies that provider was measured on`
* Missing or incorrect returned stages lower this result. Blank Ground Truth passes every vendor answer; Undisclosed Ground Truth accepts an Undisclosed or blank vendor stage.
* Each vendor is scored against the companies it was actually measured on, not a board-wide total. A vendor added later is measured on more of the enrichment cohort than one added earlier, and vendors join freshness at different snapshots, so denominators differ per row and are published alongside the score.
* “Correct when returned” divides cases that both returned a stage and were correct by every case that returned a stage. The two conditions are separate: a blank answer can still be scored correct when Ground Truth is blank or Undisclosed, so counting raw correct answers here would exceed 100%. “Latest stage returned” measures whether a provider returned a stage at all.

\[02.e\] fair execution

### Comparable inputs, provider-appropriate execution

* API providers receive the same company domain through their documented company or funding endpoint. The natural-language arms receive one shared instruction naming the same company and domain, plus the same output schema. Crunchbase is measured from a self-serve-plan export matched to the frozen cohort, not from an API request.
* Three vendors are measured on more than one endpoint, and each arm changes exactly one thing while the instruction and the output schema stay identical. Exa runs `/search` at deep-reasoning and at instant, plus its Agent API; Parallel runs the Task API and the Responses API; Seltz runs its companies and news scopes. Contract tests beside the runner assert that parity, so a difference in score is attributable to the varied parameter rather than to a different question.
* ZoomInfo and Crustdata are called in batches rather than one company at a time, so their latency is the batch round trip recorded against each company in it, not a per-company request time. Read those two rows accordingly.
* Raw responses, normalized fields, request latency, status, and failure reason are stored for every run. Only DNS/transient transport failures are retried; non-resolutions and rate-limit outcomes remain visible.
* Harmonic is measured from a reviewed export mapped to the frozen identities, but has no comparable self-serve endpoint; it is excluded from latency and cost comparisons. PitchBook and Tracxn are private-market investor-research products, including PE workflows, rather than self-serve GTM sales enrichment APIs. Usage and pricing are shown only where a provider's billing basis is known.

\[02.f\] why the board is grouped

### Three ways to answer the same question

Both boards are subdivided by how a provider produces its answer, because a single ranked list quietly compares things that are not substitutes for each other. Every provider is asked the same question about the same companies and judged by the same policy, so the rows remain directly comparable. The grouping is a reading aid, not a separate scoring rule, and nothing is excluded or weighted differently.

* **Long Running Agent APIs** dispatch an agent that researches each company at request time, reading sources and deciding what the latest round was. They are the slowest and most expensive per call, and they are priced on how much work they do.
* **Web search APIs** query a search index and extract the answer from what comes back. Fast and cheap per call, with accuracy that tracks how well a round was covered on the open web.
* **GTM data providers** return a stored record from a maintained company database. Latency and per-record cost are predictable, and coverage is set by the vendor's own ingestion rather than by the query.

The distinction is about mechanism, so Exa and Parallel each appear in two groups. Exa's Agent API and Parallel's Task API dispatch an agent that researches the company; Exa's two `/search` arms and Parallel's Responses API resolve the question against an index. The instruction and output schema sent to all of them are identical, so the endpoint is the only variable. The grouping also explains results a single list would make look contradictory, since an agent that researches on demand can lead on historical rounds and still be beaten on rounds announced last week.

\[02.g\] reading the table

### What each column means

Every result is aligned to the same company domains. A latest stage counts as correct under the documented Ground Truth matching policy, which includes selected stage equivalents and exact date or amount evidence for non-Series disagreements. A missing provider stage lowers the correct-stage result, except that a blank vendor stage is accepted when Ground Truth is Undisclosed.

* **Measured on** is that provider's own denominator: each provider/company/snapshot observation it was scored on, and how many snapshots those came from. A company repeated in a later snapshot counts again because it is a new query moment. Two rows with different values here were not measured on the same cohort.
* **Latest stage correct** is the headline. Of the scoreable companies, the share where the provider named the right stage.
* **Correct when returned** divides correct answers that actually returned a stage by every case where the provider returned one.
* **Funding fields returned** counts how often latest stage, date, round amount, total raised, and round count were populated. It does not check whether the four non-stage fields are correct.
* **Official docs** links each provider's own API, endpoint, or access documentation. These describe vendor configuration and do not influence the score.
* **Est. cost** is modeled from observed billing units at published rates, scaled to the companies that board scored.

\[03\] what is not measured+ 

\[03.a\] questions this benchmark does not ask

* **Whether the other funding fields are right.** Only the latest stage is judged for correctness with announced date as fallback. Round amount, total raised, and round count are counted for presence only, so a provider can score well on funding fields returned while carrying wrong amounts.
* **Full funding history.** One round per company is evaluated: the latest one. A provider with an excellent complete round history and a weak latest round scores badly here.
* **Investor, valuation, or board data.** Not requested and not scored, even where a provider returns it.
* **Everything else the vendor sells.** Most of these are broad platforms. Contact data, intent, technographics, and workflow tooling are different products. A result here says nothing about them.

\[03.b\] limits on the numbers reported

* **Rows do not always contain the same measurement observations.** Providers join at different times, so each is scored on its own denominator. Comparing two rows with very different _Measured on_ counts compares performance on two different cohorts.
* **Cost is modeled, not invoiced.** It is derived from observed billing units at published rates and scaled to each board. Plan minimums, negotiated rates, and credit rules will move real spend. Explorium's figure is an explicit request proxy because its per-endpoint credit burn is not public.
* **Latency is one runner on one network.** Median request time is indicative, and batch endpoints are timed per batch rather than per company.
* **Reviewed exports are not timed runs.** Crunchbase and Harmonic are mapped from supplied exports, so they carry no latency or cost and cannot be read as evidence of API behaviour.
* **Freshness is a rolling number.** Each snapshot is a different set of companies. A change between snapshots can be a real shift in a provider, or a harder cohort.

\[04\] faq

## Frequently asked questions

### Where are the official API docs for the funding data providers in this benchmark?

The provider table links directly to official API, endpoint, product, or access documentation for every measured provider, including Crunchbase, Harmonic, Apollo, People Data Labs, PredictLeads, Exa, and Parallel. Those links document vendor access and capabilities; vendor claims do not determine benchmark scores.

### Does Openbenchmarks verify funding events with official sources?

Yes. Ground Truth is independently reviewed against official company newsrooms, investor announcements, regulatory filings where applicable, and company-issued wire releases. Provider outputs are then scored against that frozen reference using identical company inputs.

### What is the difference between the freshness board and the enrichment board?

They answer different questions and are ranked separately. Freshness measures rounds announced in the trailing 30 days, so it tells you how fast a provider indexes a new announcement. Enrichment measures rounds older than 30 days, so it tells you how completely a provider has backfilled history. A provider can lead one and trail the other, which is exactly why they are not averaged into a single rank.

### Why are providers scored on different numbers of companies?

Because they were measured on different companies. Providers join the benchmark at different times, and each freshness snapshot is its own cohort that ages into enrichment once its window closes. Every score is therefore computed against that provider's own denominator, shown in the Measured on column. Using a board-wide denominator would silently penalise every provider that was present before a later one joined.

### Which company funding data provider is the most accurate?

It depends on which board you need, so read the one that matches your workflow rather than a blended rank. The tables report latest stage correct as the headline, alongside correct when returned, which separates a provider that is often wrong from one that is often silent. Both are shown because a provider that returns nothing is not the same as a provider that returns a wrong stage, and the right trade-off depends on what consumes the data.

### Are Crunchbase and Harmonic comparable to the API providers?

Only on correctness and coverage. Both are measured from reviewed exports mapped to the same company identities rather than from timed, billed endpoint calls, so neither carries a latency or cost figure. An export also has no query moment relative to an announcement, which is what a freshness score measures.

### How is a latest funding stage decided to be correct?

Against a documented matching policy, judged by gpt-5.6 at medium reasoning effort with the same prompt for every provider. The policy accepts documented equivalents for same-letter Series variants, PE and growth, strategic investment, Seed Bridge and Pre-Series A, and crowdfunding and grant labels, and keeps Seed and Pre-Seed distinct. For non-Series disagreements an exact announced date or amount match can establish correctness. Every verdict carries a reason and a decision basis.

### Can I reproduce these numbers?

Yes. The frozen inputs, the normalized provider outputs, the judge policy, and the evaluation code are public in the open data repository linked at the top of this page. Each freshness cohort publishes its own input list, so a snapshot is reproducible on its own rather than only as part of the combined cohort. Literal vendor HTTP response bodies are not redistributed.

### Why are PitchBook and Tracxn not benchmarked here?

They are private-market research platforms built primarily for investor workflows, including private equity, rather than company funding enrichment APIs that a GTM system can call per domain. This benchmark measures the enrichment job, so a provider has to expose a way to ask about a specific company and get a structured answer back.

\[05\] changelog+ 

2026-09-15 Nimble Search added

Added Nimble Search lite and standard to both funding boards: 300 companies for enrichment and 46 for the September 13 freshness snapshot. Both use `focus=general`, `full_content=false`, and the shared funding output schema.

2026-08-26 Firecrawl Spark 2 added

* Added **Firecrawl Spark 2** as a separate measured arm alongside **Firecrawl Spark 1 Mini**, using the same prompt, output schema, and agent endpoint so the model is the only changed variable.
* Rebenchmarked every vendor on the **freshness board** against a new 50-company snapshot of recently announced funding rounds.

2026-08-17 New vendors added; benchmark split into two boards

* Six new measured arms across four vendors: the **Parallel Responses API** at medium reasoning effort, the **Exa Search API** with `type=instant`, the **Exa Agent API**, **Seltz** on its `companies` and `news` scopes, and **Firecrawl**'s agent endpoint.
* Three vendors are now measured on more than one endpoint, and each arm changes exactly one thing while the instruction and output schema stay identical: Exa across `/search` at deep-reasoning and instant plus its Agent API, Parallel across `POST /v1/tasks/runs` and `POST /v1/responses`, and Seltz across its two search scopes. Contract tests beside the runner assert that parity, so a score difference is attributable to the varied parameter rather than a different question.
* The variable is priced very differently in each case. Exa charges $0.007 per instant search, $0.015 per deep-reasoning search and $0.10 per pinned agent run, which makes the accuracy-per-dollar curve inside one vendor directly readable. Parallel charges $50 per 1,000 Responses requests against $25 per 1,000 Task runs. Firecrawl bills dynamic credits per run, so its usage is measured from what each run reports rather than assumed.
* Both boards are now **subdivided by how each provider answers**: long running agent APIs that dispatch an agent to research a company at request time, web search APIs that extract from an index, and GTM data providers that return a stored record. Every row is judged identically and stays directly comparable; the grouping stops a single ranked list from implying that a two-minute research agent and a 200ms database lookup are interchangeable purchases. Exa and Parallel each appear in two groups, because an agent endpoint and a search endpoint from the same vendor work differently.
* The benchmark now publishes **two boards** instead of one blended ranking. **Freshness** measures rounds announced in the trailing 30 days, as dated snapshots. **Enrichment** measures rounds older than that. A provider that indexes fast and one that indexes deeply were previously averaged into a single number that described neither.
* Each freshness snapshot ages into the enrichment cohort once its window closes, so enrichment grows every cycle and freshness stays a rolling measurement rather than a frozen one. Added the **second freshness snapshot**, scored by the same stage policy.
* Vendors added after a freshness window has closed are measured on enrichment only for that window. Scoring them on a month-old snapshot would measure what they knew long after the announcement, which is not what a freshness score means.
* Scores are now computed on **per-provider denominators**. Providers join at different times, so each is scored against the companies it was actually measured on, shown in the _Measured on_ column. A board-wide denominator would have penalised every provider present before a late joiner arrived.
* Estimated cost is scaled to the number of companies each board scored, rather than restating the original 300-domain figure on boards that measured fewer.
* Corrected **correct when returned**, which could exceed 100%. The judge passes any vendor answer where Ground Truth is blank and accepts a blank answer where it is Undisclosed, so a case can be scored correct having returned no stage; the numerator now counts only cases that both returned and were correct. The headline correct-stage yield and the fill rate both divide by eligible cases and never changed.
* Separately, fixed a counting error in **funding field coverage**. The metric only recognised the field names emitted by the original provider adapters (`latest_date`, `round_count`), so any provider scored through the later web-research schema, which emits `latest_announced_on` and `funding_round_count`, was capped at 3 of 5 fields no matter what it actually returned. Both spellings are now counted; coverage rose for the affected providers, and no correctness score, ranking, or judge verdict changed.

2026-08-04 Harmonic added

* Added **Harmonic** as the 13th measured provider from a supplied 300-record company export, identity-audited against the frozen cohort.
* Normalized its funding fields and judged every latest-stage answer with the same `gpt-5.6` medium-reasoning policy as the rest of the benchmark: **222 / 300 correct (74.0%)**.
* Harmonic is ranked for funding-stage correctness. Because this was an export rather than a timed, billed endpoint run, latency and cost remain intentionally blank.

2026-08-01 current snapshot

* Refreshed the 300-company funding **Ground Truth** through manual review.
* Added the LLM-judged latest-stage snapshot, evaluated with `gpt-5.6` at medium reasoning effort across all 13 providers.
* Added an auditable judge decision and concise reason to every provider-stage result.
* Added documented equivalence rules for same-letter Series variants, PE/growth, strategic-investment, Seed Bridge/Pre-Series A, and crowdfunding/grant labels.
* Kept Seed and Pre-Seed distinct; for non-Series disagreements, an exact announced-date or amount match may establish correctness.
* Blank Ground Truth passes every vendor answer; Undisclosed Ground Truth accepts an Undisclosed or blank vendor stage.
* Rejudged the affected Series-plus and Seed/Pre-Seed cases, plus ZYT after its Ground Truth changed to Pre-IPO.

go deeper

## Choosing a company funding data provider

[Best company funding data API: accuracy, latency, and cost →](https://openbenchmarks.com/company-funding/best-company-funding-data-api) · [Best funding data providers: the measured landscape →](https://openbenchmarks.com/company-funding/best-company-funding-data-providers) · [Real-time funding data: recent-round freshness and latency →](https://openbenchmarks.com/company-funding/real-time-funding-data) · [Crunchbase alternatives, measured →](https://openbenchmarks.com/company-funding/crunchbase-alternatives) · [Harmonic alternatives, measured →](https://openbenchmarks.com/company-funding/harmonic-alternatives) · [PredictLeads vs People Data Labs →](https://openbenchmarks.com/company-funding/predictleads-vs-people-data-labs) · [Apollo vs People Data Labs →](https://openbenchmarks.com/company-funding/apollo-vs-people-data-labs) · [PredictLeads vs Apollo →](https://openbenchmarks.com/company-funding/predictleads-vs-apollo) · [Exa vs Parallel →](https://openbenchmarks.com/company-funding/exa-vs-parallel) 

Openbenchmarks [all benchmarks](https://openbenchmarks.com/) [crunchbase alternatives](https://openbenchmarks.com/company-funding/crunchbase-alternatives) [company enrichment](https://openbenchmarks.com/company-enrichment) © 2026 Openbenchmarks