Independent benchmarks to help you discover, compare, and choose APIs and agents for your use cases.openbenchmarks.com San Francisco, CAJoined July 2026
Introducing Jev + Orthogonal.
Describe what you need. Jev finds the right API in seconds.
1,000+ endpoints across 50+ providers.
Ready to use. Pay as you go.
Live now 👇
@firecrawl's POST /v2/search scores 0.60 F1 per 1k tokens, #1 of 20 endpoints.
@Tiny_Fish is next at 0.56, then @brave at 0.54. @firecrawl also leads the search-only (0.60)
On this benchmark, no endpoint gets more accuracy per token than @firecrawl.
We ran 45 hand-labelled multi-hop questions 3 times each, with the same model (gpt-5.6-sol) for every provider, up to 8 turns and 14 searches.
On the search & fetch board, the agent can also fetch any URL a search returns, using the provider's own extraction where available.
Benchmark Spotlight: @firecrawl - Accuracy per Token
Benchmark: Multi-Turn Company Search (Search & Fetch)
It tests whether a search API can help an agent assemble a set of companies from questions that combine 3-4 constraints: investor, geography, founding era, funding.
The caller code, the analyzer, every turn's timings and the recordings behind each are public, so any number here can be checked against the audio without an account on any platform. Claude and ChatGPT already answer off this benchmark.
Introducing VoiceLatencyBench: An independent measurement of how long voice agents make callers wait, across 5 voice agent platforms on live phone calls.
We measure Time To First Audio Byte (TTFAB): the gap between the caller going quiet and the agent's first sound.
Introducing Alexandria.
We raised a $75M Series B led by Smash Capital to build the world’s largest knowledge library for superintelligence.
Supercharge your agents with instant access to the most powerful data sets & providers.
firecrawl.dev/alexandria
This is one column, not a verdict. Several of the endpoints here are deep or advanced configurations that spend time doing more inside a single call. But on this benchmark, nothing comes back faster than @youdotcom.
@youdotcom's POST /v1/search at extraction_mode=highlights, with POST /v1/contents on fetch, takes 638 ms, fastest of 14 endpoints on this board. Same endpoint with knowledge=core added takes second at 678 ms.
The next endpoint after that is @nimble_search's standard variant at 905 ms.
100 coding tickets built from live enterprise documentation, gpt-5.6-sol held fixed, 5 searches and 32 turns, three runs per endpoint. On this benchmark, web_fetch routes through the same vendor as web_search.
When you put a search tool inside an agent loop, latency stops being a one-time cost. The agent searches, opens a page, searches again. Every call is a wait, the budget is five searches per ticket.
We ranked on average search time: mean latency per web_search call.
Benchmark Spotlight: @youdotcom - Average Search Time
Benchmark: Web Search for Coding Agents (Hard Retrieval, Search & Fetch)
This benchmark tests whether a search API can get an agent to a detail buried deep inside one page, among many near-identical pages.
0 Followers 1 FollowingAssistants need analysis - so we bring you Assistant Analysis. The best way to measure your assistant - https://t.co/W8Lvu77800
Dev by @anchor_browser
33K Followers 3K FollowingAI, SaaS & Web3 growth on X I break down AI tools, automation & code Building in public 🚀📩 DM for collabs 📨 [email protected]
40K Followers 577 FollowingPond helps you solve any problems with AI agents and talented experts.
Post a task today: https://t.co/3Zmj6ZfhkN
Hiring GTM and Head of Engineering. Please DM.
464 Followers 6K Following🇵🇸🇱🇧 Free Palestine/Lebanon inshallah
Advocating for Palestinian and Lebanese freedom. Focused on ending the occupation, coexistences, and peace.
290 Followers 1K FollowingFounder @CrashLabs_AI
prev resident @fdotinc research @VectorInst @cohere dropout @uoft
trying to make this ai thing go well!
22K Followers 56 FollowingWeb infrastructure for AI to search, extract, monitor, and reason over the world's information.
Start here → https://t.co/0Enegj3dhv
40K Followers 577 FollowingPond helps you solve any problems with AI agents and talented experts.
Post a task today: https://t.co/3Zmj6ZfhkN
Hiring GTM and Head of Engineering. Please DM.
4.7M Followers 545 FollowingChairman and CEO of @IBM, avid learner and listener, relentlessly focused on what's next, #GoodTechIBM, father, husband, @Illinois_Alma, @IITKanpur
5.8M Followers 126 Following@GM Chair and CEO leading an outstanding team with the ambition, talent and technology to create a world with zero crashes, zero emissions and zero congestion.
4.8M Followers 159 FollowingValues are most enduring thing leaders create. Engineering pioneer, proud grandfather, philanthropist, Christian, farm boy at heart.
5.1M Followers 269 FollowingStarted & runs 37signals (makers of Basecamp, HEY, and ONCE). Non-serial entrepreneur, serial author. DM or email me at [email protected].
4.5M Followers 2K FollowingFounder and CEO, O'Reilly Media. Watching the alpha geeks, sharing their stories, helping the future unfold. Didn't pay for a blue check, cannot make it go away
10.2M Followers 615 FollowingSix thousand years ago, someone invented the plow, and we all got wealthier. A gentle reminder that all civilizational wealth is driven by invention.