What we have actually tested
Every entity below is backed by executions in the archive. A ranking is a computation over them, not a document — change a finding and the answer above it changes with it.
Benchmarks
AI Customer Support Chatbots
Automate customer support using an AI chatbot / AI customer support agent — which agents handle real customer support conversations for a business best?
v1 · draft
AI Database Agents
Which AI database agent answers business questions about a live relational database correctly — in the company's own terms, and honestly when the data cannot answer?
v1 · draft
Managed RAG
Which managed RAG service turns a folder of private documents into a query API that finds the right passage, answers only from what it retrieved, cites the source that actually supports the claim — and keeps up when the source changes?
v1 · draft
Converting a complex PDF into clean Markdown with a hosted API
Which hosted API converts a complex, real-world PDF into faithful, usable Markdown?
v1 · draft
Converting a complex PDF into clean Markdown with an open-source library
Which self-hostable open-source library converts a complex, real-world PDF — native text, dense tables, charts, images and scanned pages — into Markdown a reader can trust without opening the original?
v1 · draft
Quiz Generation
Which AI quiz tool turns study material into valid quizzes, and then grades, explains and tracks what the learner actually understood?
v1 · draft
Structured Document Extraction
Which document extraction platform turns a user-defined schema into correct structured records — across layouts, scans and repeated sets — and is honest about what it could not find?
v1 · draft
Web Search for AI
Which web search or answer API finds the live-web material that actually supports an answer, keeps up when that material changes, and -- where it writes the answer itself -- says only what its own sources support?
v1 · draft
Capabilities
Authored by hand. Never derived from similarity.
Scenarios
The questions. Reusable across studies, inputs and tools.
| Scenario | Question | Grades |
|---|---|---|
| a-bad-file-in-the-upload | A bad file in the upload | 0 |
| a-claim-in-the-answer-traced-back-to-its-source | A claim in the answer traced back to its source | 0 |
| a-cleanly-scanned-document | A cleanly scanned document | 0 |
| a-configured-rule-requires-escalation | A configured rule requires escalation | 0 |
| action-failure | Requested action cannot be completed | 0 |
| action-requires-confirmation | Action requires confirmation | 2 |
| a-document-containing-a-code-block | A document containing a code block | 0 |
| a-document-containing-mathematical-equations | A document containing mathematical equations | 0 |
| a-document-mixing-digital-and-scanned-pages | A document mixing digital and scanned pages | 0 |
| a-document-with-a-data-chart | A document with a data chart | 0 |
| a-document-with-figures-and-captions | A document with figures and captions | 0 |
| a-document-with-styled-headings-and-subheadings | A document with styled headings and subheadings | 0 |
| a-field-the-tool-can-only-get-right-from-examples | A field the tool can only get right from examples | 0 |
| agent-actions-tool-calls-can-be-inspected | Agent actions/tool calls can be inspected | 0 |
| agent-cannot-resolve-the-request | Agent cannot resolve the request | 0 |
| aggregate-across-documents | Aggregate across documents | 0 |
| a-learner-takes-a-quiz-on-their-own | A learner takes a quiz on their own | 0 |
| allowlisted-web-search | Search only these websites | 0 |
| ambiguous-request | Request is unclear and needs clarification | 0 |
| analytics-accuracy | Reported analytics match what actually happened | 0 |
| a-new-page-has-just-gone-live | A new page has just gone live | 0 |
| an-excluded-column-is-asked-for-directly | An excluded column is asked for directly | 0 |
| an-excluded-table-is-needed-to-answer-the-question | An excluded table is needed to answer the question | 0 |
| an-ordinary-digital-text-document | An ordinary digital text document | 0 |
| answer-is-available-in-the-knowledge-base | Answer is available in the knowledge base | 0 |
| answer-is-not-available-in-the-knowledge-base | Answer is not available in the knowledge base | 0 |
| a-page-laid-out-in-multiple-columns | A page laid out in multiple columns | 0 |
| a-page-on-the-web-answers-the-question | A page on the web answers the question | 0 |
| a-repeated-set-of-records | A repeated set of records | 0 |
| a-simple-clearly-formatted-table | A simple, clearly formatted table | 0 |
| a-table-that-continues-across-a-page-break | A table that continues across a page break | 0 |
| a-teacher-gives-the-quiz-to-a-class | A teacher gives the quiz to a class | 0 |
| business-defined-term | The question uses a term the company defines itself | 0 |
| class-struggle-report | The teacher checks what the class struggled with | 0 |
| configured-instruction-changes-agent-behavior | Configured instruction changes agent behavior | 0 |
| configured-restriction-is-respected | Configured restriction is respected | 0 |
| conflicting-sources | The retrieved sources disagree with each other | 0 |
| content-is-removed-from-the-source | Content is removed from the source | 0 |
| conversation-activity-can-be-inspected | Conversation activity can be inspected | 0 |
| conversation-context-retention | Information from earlier in the conversation is needed later | 0 |
| correct-a-field-value | Correct a field value | 0 |
| corrected-data-goes-downstream | Corrected data goes downstream | 0 |
| customer-communicates-in-another-supported-language | Customer communicates in another supported language | 0 |
| customer-corrects-information-given-earlier | Customer corrects information given earlier | 0 |
| customer-switches-languages-during-the-conversation | Customer switches languages during the conversation | 0 |
| definition-correction | The agent used the wrong definition and the user corrects it | 0 |
| document-scoped-retrieval | Retrieval is scoped to part of the documents | 0 |
| existing-content-is-changed | Existing content is changed | 0 |
| external-system-retrieval | Retrieve information from an external system | 0 |
| failed-question-surfacing | The operator needs to find the questions the agent is failing on | 0 |
| feedback-can-be-submitted | Feedback can be submitted | 0 |
| few-shot-generalization | The user supplies example questions and the queries they trust | 0 |
| filter-records-across-documents | Filter records across documents | 0 |
| find-reported-bad-answer | A user reports a bad answer and the operator has to find it | 0 |
| follow-up-suggestions | The agent offers what to ask next | 0 |
| footnotes-at-the-bottom-of-the-page | Footnotes at the bottom of the page | 0 |
| learning-from-correction | Correction changes future behavior where the product claims learning | 0 |
| learning-goal-quiz | The user asks the quiz to test a particular learning goal | 0 |
| mixed-correct-incorrect-responses | The learner answers some fixed-answer questions correctly and others incorrectly | 0 |
| name-disambiguation | Something more famous has the same name | 0 |
| new-content-is-added-to-the-source | New content is added to the source | 0 |
| one-message-contains-multiple-requests | One message contains multiple requests | 0 |
| one-topic-is-weaker-than-the-others | One topic is weaker than the others | 0 |
| paraphrased-request | The request uses different wording than the source | 0 |
| plain-fact-or-short-list | The answer is a plain fact or a short list | 0 |
| relative-date-report | The report is run again after the data has changed | 0 |
| repair-a-record | Repair a record | 0 |
| resume-earlier-conversation | Earlier conversation/session needs to be continued | 0 |
| same-answer-different-format | The user asks to see the same answer a different way | 0 |
| source-action-disambiguation | Request requires choosing the correct source or action | 0 |
| split-across-text-and-diagram | Important information appears in both text and a diagram | 0 |
| standing-rule-persistence | The user sets a standing rule for all future answers | 0 |
| supported-answer | The retrieved material supports an answer | 0 |
| the-answer-is-available-in-the-connected-data | The answer is available in the connected data | 0 |
| the-answer-needs-a-summary-and-its-detail-together | The answer needs a summary and its detail together | 0 |
| the-answer-was-never-in-the-schema | The answer was never in the schema | 0 |
| the-connected-data-cannot-answer-the-question | The connected data cannot answer the question | 0 |
| the-databases-names-are-cryptic-and-the-user-documents-them | The database's names are cryptic and the user documents them | 0 |
| the-field-is-absent | The field is absent | 0 |
| the-follow-up-is-ambiguous | The follow-up is ambiguous | 0 |
| the-learner-answers-correctly-in-their-own-words | The learner answers correctly in their own words | 0 |
| the-learner-checks-whether-they-are-improving | The learner checks whether they are improving | 0 |
| the-learner-gives-a-clearly-wrong-answer | The learner gives a clearly wrong answer | 0 |
| the-learner-returns-to-the-material-later | The learner returns to the material later | 0 |
| the-learners-answer-is-partly-correct | The learner's answer is partly correct | 0 |
| the-next-question-refers-to-the-previous-answer | The next question refers to the previous answer | 0 |
| the-same-field-across-layouts | The same field across layouts | 0 |
| the-set-contains-a-non-record | The set contains a non-record | 0 |
| the-set-continues-across-a-page-break | The set continues across a page break | 0 |
| the-set-is-legitimately-empty | The set is legitimately empty | 0 |
| the-user-changes-direction-mid-conversation | The user changes direction mid-conversation | 0 |
| the-user-narrows-what-they-just-asked | The user narrows what they just asked | 0 |
| the-users-wording-doesnt-match-how-the-values-are-stored | The user's wording doesn't match how the values are stored | 0 |
| the-user-turns-a-recorded-lecture-into-a-quiz | The user turns a recorded lecture into a quiz | 0 |
| the-user-turns-study-material-into-a-quiz | The user turns study material into a quiz | 0 |
| the-user-wants-several-findings-in-one-place | The user wants several findings in one place | 0 |
| the-user-wants-to-keep-an-answer-they-just-got | The user wants to keep an answer they just got | 0 |
| the-value-is-directly-available | The value is directly available | 0 |
| the-value-must-be-derived | The value must be derived | 0 |
| the-value-needs-a-supplied-definition | The value needs a supplied definition | 0 |
| trace-a-value-to-its-location | Trace a value to its location | 0 |
| trend-or-comparison-form | The answer is a trend or a comparison | 0 |
| two-users-are-given-different-data-scope | Two users are given different data scope | 0 |
| unsupported-answer | The retrieved material doesn't support an answer | 0 |
| updated-known-page | A page the tool already finds has changed | 0 |
| update-something-in-an-external-system | Update something in an external system | 0 |
| user-explicitly-asks-for-a-human | User explicitly asks for a human | 0 |
| user-is-not-authorized-to-perform-the-action | User is not authorized to perform the action | 0 |
| user-question-audit | Someone needs to see what people have been asking | 0 |
Tools
| Tool | Runs | |
|---|---|---|
| Academa AI | 0 | never executed |
| Ada | 0 | never executed |
| Adobe API | 0 | never executed |
| Adobe PDF Extract API | 9 | |
| Adobe Podcast Enhance | 0 | never executed |
| Affinda | 0 | never executed |
| AICloneVoiceFree.com | 0 | never executed |
| AI for Database | 0 | never executed |
| Airparser | 0 | never executed |
| AKOOL | 0 | never executed |
| Amazon Transcribe | 0 | never executed |
| AnimG | 0 | never executed |
| Anomaly AI | 0 | never executed |
| Antigravity | 0 | never executed |
| Apollo | 0 | never executed |
| AskYourDatabase | 0 | never executed |
| AssemblyAI | 0 | never executed |
| Audo Studio | 0 | never executed |
| Auphonic | 0 | never executed |
| AutoCaption.io | 0 | never executed |
| AWS Transcribe | 0 | never executed |
| Basedash | 0 | never executed |
| BlazeSQL | 5 | |
| Boords | 0 | never executed |
| Botpress | 0 | never executed |
| Botsonic | 0 | never executed |
| Brave Search API | 7 | |
| Bria.ai | 0 | never executed |
| Camb.AI | 0 | never executed |
| camelAI | 0 | never executed |
| CapCut | 0 | never executed |
| Captions | 0 | never executed |
| Careerflow AI | 0 | never executed |
| Cartesia.ai | 0 | never executed |
| Chatbase | 0 | never executed |
| ChatGPT | 0 | never executed |
| ChatGPT Canvas | 0 | never executed |
| Claude AI | 0 | never executed |
| Claude Artifacts | 0 | never executed |
| Claude AI | 0 | never executed |
| Clay | 0 | never executed |
| Cleanvoice | 0 | never executed |
| Cognee | 0 | never executed |
| CoSupport.ai | 0 | never executed |
| Creatify.ai | 0 | never executed |
| CustomGPT.ai | 0 | never executed |
| Cutout.Pro | 0 | never executed |
| CVParserPro | 0 | never executed |
| Dante AI | 0 | never executed |
| Datalab | 0 | never executed |
| Decagon | 0 | never executed |
| Deepgram | 0 | never executed |
| Definite | 0 | never executed |
| Denser AI | 0 | never executed |
| Descript | 0 | never executed |
| Descript Studio Sound | 0 | never executed |
| D-ID | 0 | never executed |
| Dify | 0 | never executed |
| doc2mark | 6 | |
| Docling | 0 | never executed |
| Docparser | 0 | never executed |
| Docsumo | 0 | never executed |
| DoctoQuiz | 0 | never executed |
| DocTR | 0 | never executed |
| Dolphin (ByteDance) | 0 | never executed |
| Dot | 0 | never executed |
| Drawstory | 0 | never executed |
| Draxlr | 0 | never executed |
| DreamFace | 0 | never executed |
| Dubverse | 0 | never executed |
| Dust | 0 | never executed |
| EasyMotion | 0 | never executed |
| Edcafe AI | 0 | never executed |
| ElevenLabs | 0 | never executed |
| ElevenLabs Scribe | 0 | never executed |
| ElevenLabs Voice Isolator | 0 | never executed |
| Epidemic Sound | 0 | never executed |
| Epidemic Sound Voices | 0 | never executed |
| Exa | 0 | never executed |
| Extend AI | 7 | |
| Extracta.ai | 0 | never executed |
| Fathom | 0 | never executed |
| Fellow | 0 | never executed |
| Intercom Fin | 0 | never executed |
| Fine Voice | 0 | never executed |
| Firecrawl | 9 | |
| Fireflies | 0 | never executed |
| Fireflies.ai | 0 | never executed |
| Fish Audio | 0 | never executed |
| FlexClip | 0 | never executed |
| Fotor | 0 | never executed |
| Framia | 0 | never executed |
| Framia.converge.ai | 0 | never executed |
| Freshdesk Freddy | 11 | |
| FS Agent (DIY control) | 0 | never executed |
| Futuresmart AI | 0 | never executed |
| FutureSmart AI Diagram Animator | 0 | never executed |
| FutureSmart Database Agent | 21 | |
| FutureSmart Document Intelligence | 24 | |
| FutureSmart NL2SQL Agent | 0 | never executed |
| Gemini | 0 | never executed |
| Gemini Canvas | 0 | never executed |
| Gladia | 0 | never executed |
| Gling AI | 0 | never executed |
| Google Cloud Speech-to-Text | 0 | never executed |
| Google Flow | 0 | never executed |
| Gorgias | 0 | never executed |
| GPAI | 0 | never executed |
| GPT-5.4 mini | 0 | never executed |
| GPT-5.4 nano | 0 | never executed |
| GPT-5.6 luna | 0 | never executed |
| GPT-5.6 terra | 0 | never executed |
| Granola | 0 | never executed |
| Grok | 0 | never executed |
| GroqCloud | 0 | never executed |
| Gumloop | 0 | never executed |
| Hailuo | 0 | never executed |
| HappyScribe | 0 | never executed |
| Help Scout | 0 | never executed |
| HeyGen | 0 | never executed |
| Hindsight | 0 | never executed |
| Hireability | 0 | never executed |
| HrFlow | 0 | never executed |
| Hunter | 0 | never executed |
| ImagineArt | 0 | never executed |
| InVideo AI | 0 | never executed |
| Inworld | 0 | never executed |
| Inworld.ai | 0 | never executed |
| Jina | 0 | never executed |
| Jina AI Reader | 0 | never executed |
| JoggAI | 0 | never executed |
| JotForm | 0 | never executed |
| Kapwing | 0 | never executed |
| Kling AI | 0 | never executed |
| Knowt | 0 | never executed |
| Kodisc | 0 | never executed |
| Kommunicate | 15 | |
| Landing AI | 0 | never executed |
| Leonardo AI | 0 | never executed |
| Linkup | 0 | never executed |
| LiteParse | 2 | |
| LlamaParse | 4 | |
| Luma AI Dream Machine | 0 | never executed |
| MarkItDown | 7 | |
| Media.io | 0 | never executed |
| MeetGeek | 0 | never executed |
| Mem0 | 0 | never executed |
| Microsoft Azure Speech Services | 0 | never executed |
| MinerU | 0 | never executed |
| Minimax | 0 | never executed |
| MiniMax Hailuo | 0 | never executed |
| MiniMax.io | 0 | never executed |
| Mirelo | 0 | never executed |
| Mistral AI | 10 | |
| Murf AI | 0 | never executed |
| Nanonets | 16 | |
| Noise Remover | 0 | never executed |
| Noiz AI | 0 | never executed |
| NotebookLM | 0 | never executed |
| NoteGPT | 0 | never executed |
| Notta | 0 | never executed |
| Nutrient.io | 0 | never executed |
| Olostep | 0 | never executed |
| OpenAI | 0 | never executed |
| OpenAI Speech-to-Text | 0 | never executed |
| OpenAI web search | 0 | never executed |
| OpenResume | 0 | never executed |
| Opus Clip | 0 | never executed |
| Otter | 0 | never executed |
| Otter.ai | 0 | never executed |
| PaddleOCR | 0 | never executed |
| PaddleOCR-VL | 0 | never executed |
| Parseur | 0 | never executed |
| PDF.ai | 0 | never executed |
| PDF to Quiz | 0 | never executed |
| PDFVector | 0 | never executed |
| PDF Vector | 0 | never executed |
| Perplexity Sonar | 0 | never executed |
| Pickaxe | 0 | never executed |
| Picsart | 0 | never executed |
| Pika Labs | 0 | never executed |
| PixVerse AI | 0 | never executed |
| Premiere Pro | 0 | never executed |
| PyMuPDF4LLM | 6 | |
| Querio | 0 | never executed |
| QuestionWell | 0 | never executed |
| Quizgecko | 0 | never executed |
| Quizizz | 0 | never executed |
| QuizRise | 0 | never executed |
| Rask AI | 0 | never executed |
| Reducto | 11 | |
| Relevance AI | 0 | never executed |
| RemotionAI | 0 | never executed |
| Remotion AI | 0 | never executed |
| RemotionVideo | 0 | never executed |
| Replit | 0 | never executed |
| Respond | 0 | never executed |
| Respond.io | 0 | never executed |
| ResumeParser | 0 | never executed |
| Retab | 0 | never executed |
| Rev AI | 0 | never executed |
| revid.ai | 0 | never executed |
| Rezi AI | 0 | never executed |
| Saleshandy | 0 | never executed |
| Scenario | 0 | never executed |
| SerpAPI | 0 | never executed |
| Serper | 0 | never executed |
| Skima AI | 0 | never executed |
| Skyvern | 0 | never executed |
| Snov.io | 0 | never executed |
| Speechify | 0 | never executed |
| Speechmatics | 0 | never executed |
| Spider | 0 | never executed |
| Steve AI | 0 | never executed |
| StudyFetch | 0 | never executed |
| Studyglen | 0 | never executed |
| StudyX AI | 0 | never executed |
| Summarize.tech | 0 | never executed |
| Supermemory | 0 | never executed |
| Sync Labs | 0 | never executed |
| Synthesia | 0 | never executed |
| Synthesia.io | 0 | never executed |
| Tavily | 8 | |
| Teal | 0 | never executed |
| Tensorlake | 0 | never executed |
| Tidio Lyro | 0 | never executed |
| TopMediai Voice Cloning | 0 | never executed |
| Topview AI | 0 | never executed |
| Uberduck | 0 | never executed |
| Unstract | 0 | never executed |
| Unstructured | 0 | never executed |
| Upstage AI | 0 | never executed |
| Valyu | 0 | never executed |
| Vectara | 10 | |
| Vectorize.io | 15 | |
| VEED | 0 | never executed |
| Vidnoz AI | 0 | never executed |
| Vidyo.ai | 0 | never executed |
| Vismo Studio | 0 | never executed |
| Vizard | 0 | never executed |
| Vizard.ai | 0 | never executed |
| VocalAI | 0 | never executed |
| Voiceflow | 0 | never executed |
| Wonderchat | 0 | never executed |
| Wonsulting AI | 0 | never executed |
| X-Pilot | 0 | never executed |
| You.com | 0 | never executed |
| YouLearn AI | 0 | never executed |
| Zapier AI Agents | 0 | never executed |
| Zendesk AI | 0 | never executed |
| Zep | 0 | never executed |
| Zubtitle | 0 | never executed |