# \# \## The highest accuracy web search for your AI [Start buildingP](https://platform.parallel.ai/)[\[Start building\](https://platform.parallel.ai/)](https://platform.parallel.ai/)[Onboard your AgentA](https://docs.parallel.ai/getting-started/overview#onboard-your-agent)[\[Onboard your Agent\](https://docs.parallel.ai/getting-started/overview#onboard-your-agent)](https://docs.parallel.ai/getting-started/overview#onboard-your-agent) Built for frontier teams, trusted by Fortune 500 ![](/_next/image?url=%2Flogos%2Fgray%2Fformation-bio.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fformation-bio.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fattio.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fattio.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fgranola.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgranola.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fmonaco.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmonaco.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fmanus.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmanus.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fhex.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fhex.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fdropbox.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fdropbox.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fowner.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fowner.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fgreptile.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgreptile.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Frogo.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Frogo.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fgumloop.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgumloop.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Factively.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Factively.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Famp.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Famp.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fmacroscope.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmacroscope.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fmiddesk.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmiddesk.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fnooks.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fnooks.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fopenrouter.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fopenrouter.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fgray%2Fredis.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fredis.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) [![Harvey logo](/_next/image?url=%2Flogos%2Fgray%2Fharvey.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fharvey.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)Case Study](/blog/case-study-harvey) ![Formation Bio logo](/_next/image?url=%2Flogos%2Fgray%2Fformation-bio.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fformation-bio.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Formation Bio logo](/_next/image?url=%2Flogos%2Fgray%2Fformation-bio.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fformation-bio.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Attio logo](/_next/image?url=%2Flogos%2Fgray%2Fattio.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fattio.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Attio logo](/_next/image?url=%2Flogos%2Fgray%2Fattio.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fattio.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) [![Starbridge logo](/_next/image?url=%2Flogos%2Fgray%2Fstarbridge.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fstarbridge.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)Case Study](/blog/case-study-starbridge) ![Granola logo](/_next/image?url=%2Flogos%2Fgray%2Fgranola.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgranola.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Granola logo](/_next/image?url=%2Flogos%2Fgray%2Fgranola.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgranola.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Monaco logo](/_next/image?url=%2Flogos%2Fgray%2Fmonaco.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmonaco.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Monaco logo](/_next/image?url=%2Flogos%2Fgray%2Fmonaco.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmonaco.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Manus logo](/_next/image?url=%2Flogos%2Fgray%2Fmanus.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmanus.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Manus logo](/_next/image?url=%2Flogos%2Fgray%2Fmanus.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmanus.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Hex logo](/_next/image?url=%2Flogos%2Fgray%2Fhex.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fhex.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Hex logo](/_next/image?url=%2Flogos%2Fgray%2Fhex.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fhex.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) [![Modal logo](/_next/image?url=%2Flogos%2Fgray%2Fmodal.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fmodal.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)Case Study](/blog/case-study-modal) ![Dropbox logo](/_next/image?url=%2Flogos%2Fgray%2Fdropbox.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fdropbox.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Dropbox logo](/_next/image?url=%2Flogos%2Fgray%2Fdropbox.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fdropbox.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Owner logo](/_next/image?url=%2Flogos%2Fgray%2Fowner.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fowner.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Owner logo](/_next/image?url=%2Flogos%2Fgray%2Fowner.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fowner.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Greptile logo](/_next/image?url=%2Flogos%2Fgray%2Fgreptile.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgreptile.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Greptile logo](/_next/image?url=%2Flogos%2Fgray%2Fgreptile.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fgreptile.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Rogo logo](/_next/image?url=%2Flogos%2Fgray%2Frogo.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Frogo.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) ![Rogo logo](/_next/image?url=%2Flogos%2Fgray%2Frogo.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Frogo.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix) [![Profound logo](/_next/image?url=%2Flogos%2Fgray%2Fprofound.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fprofound.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)Case Study](/blog/case-study-profound)[![Opendoor logo](/_next/image?url=%2Flogos%2Fgray%2Fopendoor.png&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)![](/_next/image?url=%2Flogos%2Fcolor%2Fopendoor.webp&w=640&q=75&dpl=dpl_AFN1CRJM5pR2VkAsskmE2YvAw9ix)Case Study](/blog/case-study-opendoor) ## ## A web API purpose-built for AIs Powering millions of daily requests ### Highest accuracy Production-ready outputs built on cross-referenced facts, with minimal hallucination. ### Predictable costs Flex compute budget based on task complexity. Pay per query, not per token. ### Evidence-based outputs Verifiability and provenance for every atomic output. ### Trusted SOC-II Type 2 Certified, trusted by leading startups and enterprises. ## Highest accuracy at every price point State of the art across the most challenging benchmarks [View All BenchmarksB](/benchmarks)[\[View All Benchmarks\](https://parallel.ai/benchmarks)](/ai/benchmarks) Humanity's Last ExamFRAMESWebWalkerFreshQACoding 310,50PARALLEL BASIC58% / 451CPMPARALLEL ADVANCED56% / 315CPMEXA57% / 522CPMTAVILY54% / 538CPM COST (CPM) ACCURACY (%) Loading chart... CPM: USD per 1000 requests. Cost is shown on a Log scale. Parallel Others Benchmark comparison across Cost (CPM) and Accuracy (%). CPM: USD per 1000 requests. Cost is shown on a Log scale. **\*\*Dataset\*\*** We evaluated search providers against five open benchmarks covering complementary aspects of agentic search: BrowseComp (hard multi-hop questions that require navigating the live web), Frames (multi-document factoid reasoning), FreshQA (time-sensitive questions where the correct answer depends on recent web information), HLE (Humanity's Last Exam — expert-level academic questions spanning math, science, and humanities), SealQA (ambiguity-robust factoid QA with intentionally misleading snippets), WebWalker (tasks designed around following links across pages to find an answer). **\*\*Evaluation methodology\*\*** Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to `` `MAX_TOOL_CALLS=25` `` tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4\. We report accuracy of the final answer. We measure accuracy and overall cost, which includes LLM token costs and tool call costs. **\*\*Testing dates\*\*** April 19-21, 2026 ## ## Highest accuracy at every price point State of the art across the most challenging benchmarks [View All BenchmarksB](/benchmarks)[\[View All Benchmarks\](https://parallel.ai/benchmarks)](/ai/benchmarks) ## ### Humanity's Last Exam | Series | Model | Cost (CPM) | Accuracy (%) | | -------- | ----------------- | ---------- | ------------ | | Parallel | Parallel Basic | 451 | 58 | | Parallel | Parallel Advanced | 315 | 56 | | Others | Exa | 522 | 57 | | Others | Tavily | 538 | 54 | CPM: USD per 1000 requests. Cost is shown on a Log scale. **\*\*Dataset\*\*** We evaluated search providers against five open benchmarks covering complementary aspects of agentic search: BrowseComp (hard multi-hop questions that require navigating the live web), Frames (multi-document factoid reasoning), FreshQA (time-sensitive questions where the correct answer depends on recent web information), HLE (Humanity's Last Exam — expert-level academic questions spanning math, science, and humanities), SealQA (ambiguity-robust factoid QA with intentionally misleading snippets), WebWalker (tasks designed around following links across pages to find an answer). **\*\*Evaluation methodology\*\*** Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to `` `MAX_TOOL_CALLS=25` `` tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4\. We report accuracy of the final answer. We measure accuracy and overall cost, which includes LLM token costs and tool call costs. **\*\*Testing dates\*\*** April 19-21, 2026 ## ### FRAMES | Series | Model | Cost (CPM) | Accuracy (%) | | -------- | ----------------- | ---------- | ------------ | | Parallel | Parallel Advanced | 93 | 87 | | Parallel | Parallel Basic | 165 | 84 | | Others | Exa | 169 | 87 | | Others | Tavily | 189 | 83 | CPM: USD per 1000 requests. Cost is shown on a Log scale. **\*\*Dataset\*\*** We evaluated search providers against five open benchmarks covering complementary aspects of agentic search: BrowseComp (hard multi-hop questions that require navigating the live web), Frames (multi-document factoid reasoning), FreshQA (time-sensitive questions where the correct answer depends on recent web information), HLE (Humanity's Last Exam — expert-level academic questions spanning math, science, and humanities), SealQA (ambiguity-robust factoid QA with intentionally misleading snippets), WebWalker (tasks designed around following links across pages to find an answer). **\*\*Evaluation methodology\*\*** Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to `` `MAX_TOOL_CALLS=25` `` tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4\. We report accuracy of the final answer. We measure accuracy and overall cost, which includes LLM token costs and tool call costs. **\*\*Testing dates\*\*** April 19-21, 2026 ## ### WebWalker | Series | Model | Cost (CPM) | Accuracy (%) | | -------- | ----------------- | ---------- | ------------ | | Others | Exa | 210 | 74 | | Others | Tavily | 202 | 71 | | Parallel | Parallel Advanced | 101 | 73 | | Parallel | Parallel Basic | 155 | 71 | CPM: USD per 1000 requests. Cost is shown on a Log scale. **\*\*Dataset\*\*** We evaluated search providers against five open benchmarks covering complementary aspects of agentic search: BrowseComp (hard multi-hop questions that require navigating the live web), Frames (multi-document factoid reasoning), FreshQA (time-sensitive questions where the correct answer depends on recent web information), HLE (Humanity's Last Exam — expert-level academic questions spanning math, science, and humanities), SealQA (ambiguity-robust factoid QA with intentionally misleading snippets), WebWalker (tasks designed around following links across pages to find an answer). **\*\*Evaluation methodology\*\*** Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to `` `MAX_TOOL_CALLS=25` `` tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4\. We report accuracy of the final answer. We measure accuracy and overall cost, which includes LLM token costs and tool call costs. **\*\*Testing dates\*\*** April 19-21, 2026 ## ### FreshQA | Series | Model | Cost (CPM) | Accuracy (%) | | -------- | ----------------- | ---------- | ------------ | | Parallel | Parallel Advanced | 49 | 79 | | Parallel | Parallel Basic | 90 | 77 | | Others | Exa | 84 | 78 | | Others | Tavily | 89 | 78 | CPM: USD per 1000 requests. Cost is shown on a Log scale. **\*\*Dataset\*\*** We evaluated search providers against five open benchmarks covering complementary aspects of agentic search: BrowseComp (hard multi-hop questions that require navigating the live web), Frames (multi-document factoid reasoning), FreshQA (time-sensitive questions where the correct answer depends on recent web information), HLE (Humanity's Last Exam — expert-level academic questions spanning math, science, and humanities), SealQA (ambiguity-robust factoid QA with intentionally misleading snippets), WebWalker (tasks designed around following links across pages to find an answer). **\*\*Evaluation methodology\*\*** Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to `` `MAX_TOOL_CALLS=25` `` tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4\. We report accuracy of the final answer. We measure accuracy and overall cost, which includes LLM token costs and tool call costs. **\*\*Testing dates\*\*** April 19-21, 2026 ## ### Coding | Series | Model | Cost (CPM) | Accuracy (%) | | -------- | ----------------- | ---------- | ------------ | | Parallel | Parallel Advanced | 154 | 82 | | Parallel | Parallel Basic | 269 | 81 | | Others | Exa | 331 | 80 | | Others | Tavily | 352 | 75 | CPM: USD per 1000 requests. Cost is shown on a Log scale. **\*\*Dataset\*\*** A proprietary coding dataset derived from production queries to Parallel’s search API. **\*\*Evaluation methodology\*\*** Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to `` `MAX_TOOL_CALLS=25` `` tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4\. We report the accuracy of the final answer. We measure accuracy and overall cost, which includes LLM token costs and tool call costs. **\*\*Testing dates\*\*** April 19-21, 2026 ## Search, built for AIs The most accurate search tool, to bring web context to your AI agents [Give AIs Search](https://platform.parallel.ai/play/search)[\[Give AIs Search\](https://platform.parallel.ai/play/search)](https://platform.parallel.ai/play/search) SthenP [Learn More](https://parallel.ai/products/search)[\[Learn More\](https://parallel.ai/products/search)](/ai/products/search) L ## The most accurate deep and wide research Run deeper and more accurate research at scale, for the same compute budget [Run a query ](https://platform.parallel.ai/play/deep-research)[\[Run a query \](https://platform.parallel.ai/play/deep-research)](https://platform.parallel.ai/play/deep-research) DthenP Starting research... ## Build a dataset from the web Define your search criteria in natural language, and get back a structured table of matches [Create a dataset](https://platform.parallel.ai/play/find-all)[\[Create a dataset\](https://platform.parallel.ai/play/find-all)](https://platform.parallel.ai/play/find-all) FthenP ID 1 2 3 4 5 6 Entities ## Custom web enrichment Bring existing data, define output columns to research, and get fresh web enrichments back [Enrich your data](https://platform.parallel.ai/play)[\[Enrich your data\](https://platform.parallel.ai/play)](https://platform.parallel.ai/play) TthenP ID Entities Product releases SOC 2 status 1 2 3 4 5 6 7 ## Monitor any event on the web Continuously monitor for any changes on the web [Monitor the web](https://platform.parallel.ai/play/monitor)[\[Monitor the web\](https://platform.parallel.ai/play/monitor)](https://platform.parallel.ai/play/monitor) MthenP New breakthroughs in AI research Start building ## ## Towards a programmatic web for AIs Parallel is building new interfaces, infrastructure, and business models for AIs to work with the web [Try it for freeP](https://platform.parallel.ai)[\[Try it for free\](https://platform.parallel.ai)](https://platform.parallel.ai)[Join us C](/careers)[\[Join us \](https://parallel.ai/careers)](/ai/careers) ## Agent onboarding prompt Use curl to read [\[parallel.ai/agents.md\](https://parallel.ai/agents.md)](/ai/agents.md) and perform the setup to install Parallel ### ## Agent onboarding prompt Use curl to read [parallel.ai/agents.md](/agents.md) and perform the setup to install Parallel [![API interface screenshot showing search query input and structured data output](/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F884549a6d3f5da8419ef5be6c0013d17c1d722bf-1586x580.png&w=3840&q=75)](https://docs.parallel.ai/home) Deep research API interface for ChatGPT and AI agents. Enterprise-grade deep research with up to 48% accuracy vs GPT-4's 1%. Built for ChatGPT deep research assistants and complex multi-hop AI workflows. Latest updates June 9 [\[OpenClaw now has free, LLM-optimized web search by default powered by Parallel\](https://parallel.ai/blog/free-web-search-openclaw)](/blog/free-web-search-openclaw) Tags:[Partnership](/blog?tag=partnership) May 19 [\[Introducing Index by Parallel\](https://parallel.ai/blog/introducing-index-by-parallel)Today we're launching Index, a platform that helps content owners understand how AI agents use their work and earn compensation tied to the value they contribute.](/blog/introducing-index-by-parallel) Tags:[Product Release](/blog?tag=product-release) May 6 [\[Parallel Monitor API: New processor tiers, snapshots and event streams, and Basis on every event\](https://parallel.ai/blog/monitor-api)The Monitor API is now generally available, with significant upgrades, bringing us closer to a web that continuously pushes changes to power proactive agents.](/blog/monitor-api) Tags:[Product Release](/blog?tag=product-release) April 29 [\[Parallel Raises at $2 Billion Valuation to Scale Web Infrastructure for Agents\](https://parallel.ai/blog/series-b)Sequoia Capital led the round. Andrew Reed, Partner at Sequoia, joins our board. Existing investors Kleiner Perkins, Index Ventures, Khosla Ventures, First Round Capital, Spark Capital, and Terrain Capital all increased their participation. ](/blog/series-b) Tags:[Fundraise](/blog?tag=fundraise)