Exa Alternatives for Research Agents, 2026
Parallel basic leads 11 tested Exa alternatives at 46.5% F1 on 45 company search tasks, independently benchmarked in 2026.
What this page compares. This page ranks measured results from 2 boards: multi-turn company search (search only) and multi-turn company search (search and page fetch). Exa was measured and is left out of its own alternatives ranking.
How to read the result. The table ranks every complete run by its score on the same questions. Latency and cost are shown beside it and not blended into the score.
Multi-turn company search, search only, ranked by F1
Every row ran the same 45 questions. Provider names link to the official docs for the configuration measured.
| Rank | Provider | F1 | Precision | Recall | Median task time | Median cost / task | Median task tokens | List price |
|---|---|---|---|---|---|---|---|---|
| 1 | Parallel basicparallel-basic | 46.5% | 88.7% | 34.4% | 67.6 s | $1.12 | 303,467 | $5.00 / 1k queries |
| 2 | Parallel advancedparallel-advanced | 44.2% | 87.6% | 32.0% | 83.4 s | $0.62 | 159,733 | $5.00 / 1k queries |
| 3 | Linkup fastlinkup-fast | 41.1% | 82.8% | 30.3% | 64.2 s | $0.92 | 259,642 | $5.00 / 1k queries |
| 4 | Tavily advancedtavily-advanced | 41.1% | 83.7% | 29.8% | 92.2 s | $1.03 | 237,706 | $16.00 / 1k queries |
| 5 | Linkup standardlinkup-standard | 40.6% | 84.0% | 29.7% | 72.3 s | $0.94 | 249,382 | $5.00 / 1k queries |
| 6 | Youextraction_mode=highlights | 38.6% | 79.3% | 28.3% | 48.2 s | $0.97 | 264,016 | $5.00 / 1k queries |
| 7 | Youextraction_mode=highlights · knowledge=core | 38.1% | 80.9% | 27.4% | 47.6 s | $0.90 | 244,245 | $5.00 / 1k queries |
| 8 | Parallel fastparallel-fast | 38.0% | 79.9% | 27.5% | 53.9 s | $0.46 | 119,953 | $1.00 / 1k queries |
| 9 | Perplexitysearch_context_size=low | 37.8% | 79.3% | 26.8% | 48.9 s | $0.33 | 66,928 | $5.00 / 1k queries |
| 10 | Tavily basictavily-basic | 36.5% | 82.5% | 25.7% | 63.0 s | $0.61 | 142,692 | $8.00 / 1k queries |
| 11 | Parallel turboparallel-turbo | 34.7% | 80.0% | 24.8% | 46.4 s | $0.42 | 110,844 | $1.00 / 1k queries |
| 12 | Nimblesearch_depth=standard · full_content=false · focus=general | 30.7% | 70.3% | 21.3% | 53.0 s | $0.47 | 109,000 | $5.00 / 1k queries |
| 13 | Firecrawlfirecrawl | 30.4% | 77.3% | 20.7% | 75.0 s | $0.28 | 50,596 | $5.00 / 1k queries |
| 14 | Brave Searchbrave | 28.0% | 66.9% | 19.3% | 43.5 s | $0.27 | 49,005 | $5.00 / 1k queries |
| 15 | TinyFishtinyfish | 26.6% | 64.7% | 17.9% | 64.2 s | $0.21 | 51,091 | $0 / 1k queries |
| 16 | Nimblesearch_depth=lite · full_content=false · focus=general | 24.1% | 65.2% | 16.1% | 67.2 s | $0.22 | 50,682 | $1.10 / 1k queries |
| 17 | Seltzseltz-companies | 14.5% | 40.0% | 9.4% | 55.2 s | $1.75 | 500,633 | $5.00 / 1k queries |
| 18 | SERP (RapidAPI)serp | 0.4% | 0.7% | 0.3% | 31.4 s | $0.10 | 10,890 | $3.00 / 1k queries |
Multi-turn company search, search and page fetch, ranked by F1
| Rank | Provider | F1 | Precision | Recall | Median task time | Median cost / task | Median task tokens | List price |
|---|---|---|---|---|---|---|---|---|
| 1 | Perplexitysearch_context_size=high | 46.6% | 87.7% | 34.7% | 53.9 s | $0.50 | 121,239 | $5.00 / 1k queries |
| 2 | Parallel basicparallel-basic | 42.3% | 81.3% | 31.3% | 68.6 s | $1.09 | 304,826 | $5.00 / 1k queries |
| 3 | Parallel advancedparallel-advanced | 42.2% | 87.6% | 30.1% | 80.9 s | $0.60 | 152,116 | $5.00 / 1k queries |
| 4 | Linkup standardlinkup-standard | 42.0% | 90.7% | 30.5% | 81.0 s | $0.91 | 254,078 | $5.00 / 1k queries |
| 5 | Tavily advancedtavily-advanced | 41.0% | 89.4% | 29.1% | 92.6 s | $0.90 | 208,974 | $16.00 / 1k queries |
| 6 | Linkup fastlinkup-fast | 39.9% | 85.3% | 28.6% | 68.0 s | $0.90 | 246,395 | $5.00 / 1k queries |
| 7 | Parallel fastparallel-fast | 39.3% | 82.3% | 28.2% | 55.1 s | $0.44 | 119,278 | $1.00 / 1k queries |
| 8 | Youextraction_mode=highlights | 38.5% | 85.2% | 26.9% | 45.6 s | $0.86 | 241,834 | $5.00 / 1k queries |
| 9 | Youextraction_mode=highlights · knowledge=core | 38.5% | 78.2% | 28.3% | 48.5 s | $0.89 | 250,679 | $5.00 / 1k queries |
| 10 | Tavily basictavily-basic | 38.1% | 81.2% | 27.8% | 62.0 s | $0.61 | 141,892 | $8.00 / 1k queries |
| 11 | Parallel turboparallel-turbo | 36.0% | 83.6% | 25.0% | 48.1 s | $0.41 | 108,358 | $1.00 / 1k queries |
| 12 | Firecrawlfirecrawl | 33.2% | 83.3% | 22.7% | 82.1 s | $0.29 | 55,069 | $5.00 / 1k queries |
| 13 | Nimblesearch_depth=standard · full_content=false · focus=general · extract formats=[markdown] | 31.3% | 73.2% | 21.7% | 59.2 s | — | 117,345 | $5.00 / 1k queries |
| 14 | TinyFishtinyfish | 30.2% | 70.9% | 20.9% | 63.4 s | $0.22 | 53,654 | $0 / 1k queries |
| 15 | Brave Searchbrave | 29.4% | 73.5% | 20.4% | 45.1 s | $0.28 | 54,482 | $5.00 / 1k queries |
| 16 | Nimblesearch_depth=lite · full_content=false · focus=general · extract formats=[markdown] | 25.8% | 67.2% | 17.7% | 72.4 s | — | 55,416 | $1.10 / 1k queries |
| 17 | Seltzseltz-companies | 16.3% | 49.5% | 10.2% | 60.1 s | $1.74 | 502,391 | $5.00 / 1k queries |
| 18 | SERP (RapidAPI)serp | 0.0% | 0.0% | 0.0% | 33.4 s | $0.10 | 11,051 | $3.00 / 1k queries |
How the ranking was measured
- A research agent searches over several turns to list every company that matches a set of constraints, on the same questions for every API.
- Answers are scored against a fixed list of correct companies: precision, recall and F1.
- Search-only rows combine search results; search-and-fetch rows can also read pages.
- Median task time, task tokens and median cost per task are reported beside the score.
- The runner and published data are open in openbenchmarks-labs/multi-turn-company-search ↗ and openbenchmarks/OB-Company-Websearch ↗. The full methodology is on the Multi-Turn Company Search Benchmark →
Questions answered by this comparison
Which provider leads exa Alternatives for Research Agents?
Parallel basic leads this table at 46.5% F1, on 45 company search tasks.
How was this measured?
A research agent searches over several turns to list every company that matches a set of constraints, on the same questions for every API. Answers are scored against a fixed list of correct companies: precision, recall and F1. Search-only rows combine search results; search-and-fetch rows can also read pages. Median task time, task tokens and median cost per task are reported beside the score.
How should speed and cost be read beside the score?
Parallel basic recorded 67.6 s median task time and $1.12 median cost per task. The most accurate configuration is rarely the fastest or the cheapest, so each is reported in its own column.
Why is Exa not in the ranking?
Exa was measured, and its configurations are left out so they cannot rank as their own alternative. Its results are on the full benchmark page.
What does this page not test?
This page ranks measured results from 2 boards: multi-turn company search (search only) and multi-turn company search (search and page fetch). Exa was measured and is left out of its own alternatives ranking.
Related pages
- Multi-Turn Company Search Benchmark: full leaderboard and methodology →
- Linkup Alternatives for Research Agents
- Firecrawl Alternatives for Research Agents
- Best Web Search API for Company Search
- Best Search API for Research Agents
- Search API with the Best Recall
- Every multi-turn company search comparison →