Then click the Experiment button on the top-right, and select the models in the dropdown.
Enter the Prompt Template.Note that you can reference the dataset columns as variables in the prompt template using {{.
When you run the experiment, Athina will run the prompt template and generate a new dataset with the results for each model.Athina will create or overwrite a column called response in each dataset.
Now if you run evaluations, they will be run on all rows of all the datasets at once.Once the evaluations are complete, you can scroll to the right, and you will see the evaluation metrics for each eval and each dataset.
Click on the evaluation metric to see the detailed results.Learn more here.
{"docsConfig":{"$schema":"https://mintlify.com/docs.json","theme":"mint","name":"Athina","colors":{"primary":"#1976D2","light":"#419BF4","dark":"#1976D2"},"favicon":"/logo/light.svg","logo":{"light":"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/logo/light.svg","dark":"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/logo/dark.svg"},"api":{"mdx":{"server":null,"auth":{"method":"key","name":"athina-api-key"}}},"navbar":{"links":[{"label":"AI Hub","href":"https://hub.athina.ai"},{"label":"Website","href":"https://athina.ai"}],"primary":{"type":"button","label":"Sign-up","href":"https://app.athina.ai/"}},"footer":{"socials":{"twitter":"https://twitter.com/AthinaAI","linkedin":"https://linkedin.com/company/athina-ai"}},"integrations":{"ga4":{"measurementId":"G-KZVC5R2GB1"}},"navigation":{"pages":[]}},"deployment":{"deploymentMetadataValue":{"subdomain":"athinaai","hasAudioNarration":false,"actualSubdomain":"athinaai","customDomains":["docs.athina.ai"],"gitSource":{"type":"github","owner":"athina-ai","repo":"docs","deployBranch":"main","contentDirectory":"","isPrivate":false,"mountPath":"","label":""},"feedback":{"thumbs":false},"entitlements":{},"buildId":"6ae91e667acc70e71f351389:success","clientVersion":"0.0.4111","searchFilterCounts":[{"language":"default","tag":"Documentation","count":166},{"language":"default","tag":"Docs","count":79},{"language":"default","tag":"API / SDK Reference","count":54},{"language":"default","tag":"Evals","count":52},{"language":"default","tag":"Datasets","count":37},{"language":"default","tag":"Guides","count":23},{"language":"default","tag":"Logging","count":21},{"language":"default","tag":"Preset Evals","count":15},{"language":"default","tag":"Prompts","count":14},{"language":"default","tag":"Logging LLM Inferences","count":13}]},"assistantConfig":{},"basePath":"","hasChatPermissions":false,"hasWebMcpPermissions":false,"toggles":[],"googleFontsInlined":false},"customCss":[],"page":{"pagePath":"guides/datasets/comparing-models","slug":"/guides/datasets/comparing-models","pageMetadata":{"title":"Run an Experiment to compare and evaluate responses from different models","sidebarTitle":"Compare and Evaluate Multiple Models","description":null,"href":"/guides/datasets/comparing-models"},"pageType":"default","mdx":{"mdxSource":{"compiledSource":"\"use strict\";const{Fragment:_Fragment,jsx:_jsx,jsxs:_jsxs}=arguments[0];const{useMDXComponents:_provideComponents}=arguments[0];function _createMdxContent(props){const _components={a:`a`,code:`code`,em:`em`,li:`li`,ol:`ol`,p:`p`,strong:`strong`,..._provideComponents(),...props.components},{Heading,Tip}=_components;if(!Heading)_missingMdxReference(`Heading`,true);if(!Tip)_missingMdxReference(`Tip`,true);return _jsxs(_Fragment,{children:[_jsx(Tip,{children:_jsxs(_components.p,{children:[`Another way to do this is to create multiple `,_jsx(_components.a,{href:`/datasets/dynamic-columns/dynamic-columns`,children:`dynamic\ncolumns`}),` - one for each model that\nyou want to compare. `,_jsx(_components.a,{href:`/datasets/dynamic-columns/dynamic-columns-run-prompt`,children:`Learn\nmore`}),`.`]})}),`\n`,_jsx(_components.p,{children:`You can run an experiment on Athina to compare multiple models in a few clicks.`}),`\n`,_jsx(Heading,{level:`3`,id:`create-or-import-a-dataset`,children:`Create or Import a Dataset`}),`\n`,_jsxs(_components.p,{children:[`Start with a dataset. Here are some instructions on how you can `,_jsx(_components.a,{href:`/datasets/create-dataset`,children:`create a dataset`}),` on Athina.`]}),`\n`,_jsx(Heading,{level:`3`,id:`configure-experiment`,children:`Configure Experiment`}),`\n`,_jsx(`video`,{src:`https://info.athina.ai/videos/experiment-re-run.mp4`,poster:`https://info.athina.ai/videos/poster/experiment-re-run.png`,muted:true,controls:true}),`\n`,_jsxs(_components.ol,{children:[`\n`,_jsxs(_components.li,{children:[`\n`,_jsxs(_components.p,{children:[`Then click the `,_jsx(_components.strong,{children:`Experiment`}),` button on the top-right, and select the models in the dropdown.`]}),`\n`]}),`\n`,_jsxs(_components.li,{children:[`\n`,_jsx(_components.p,{children:`Enter the Prompt Template.`}),`\n`,_jsx(_components.p,{children:_jsxs(_components.em,{children:[`Note that you can reference the dataset columns as variables in the prompt template using `,_jsx(_components.code,{children:`{{`}),`.`]})}),`\n`]}),`\n`,_jsxs(_components.li,{children:[`\n`,_jsx(_components.p,{children:`When you run the experiment, Athina will run the prompt template and generate a new dataset with the results for each model.`}),`\n`,_jsx(_components.p,{children:_jsxs(_components.em,{children:[`Athina will create or overwrite a column called `,_jsx(_components.code,{children:`response`}),` in each dataset.`]})}),`\n`]}),`\n`]}),`\n`,_jsx(Heading,{level:`3`,id:`compare-responses`,children:`Compare Responses`}),`\n`,_jsxs(_components.p,{children:[`Athina will also open all these datasets in the “Datasets” tab, so you can `,_jsx(_components.a,{href:`/datasets/compare-datasets`,children:`compare them side-by-side`}),`.`]}),`\n`,_jsx(`video`,{src:`https://info.athina.ai/videos/compare-responses-diff-view.mp4`,poster:`https://info.athina.ai/videos/poster/compare-responses-diff-view.png`,muted:true,controls:true}),`\n`,_jsx(Tip,{children:_jsx(_components.p,{children:`You can turn the diff view on or off by clicking the “View Options” button on\nthe top-right.`})}),`\n`,_jsx(Heading,{level:`3`,id:`run-evaluations`,children:`Run Evaluations`}),`\n`,_jsx(_components.p,{children:`Now if you run evaluations, they will be run on all rows of all the datasets at once.`}),`\n`,_jsx(_components.p,{children:`Once the evaluations are complete, you can scroll to the right, and you will see the evaluation metrics for each eval and each dataset.`}),`\n`,_jsx(`video`,{src:`https://info.athina.ai/videos/run-evals-on-dataset.mp4`,poster:`https://info.athina.ai/videos/poster/run-evals-on-dataset.png`,controls:true,muted:true}),`\n`,_jsx(Heading,{level:`3`,id:`view-the-evaluation-metrics`,children:`View the Evaluation Metrics`}),`\n`,_jsx(_components.p,{children:`Click on the evaluation metric to see the detailed results.`}),`\n`,_jsxs(_components.p,{children:[`Learn more `,_jsx(_components.a,{href:`/datasets/metrics`,children:`here`}),`.`]})]})}function MDXContent(props={}){const{wrapper:MDXLayout}={..._provideComponents(),...props.components};return MDXLayout?_jsx(MDXLayout,{...props,children:_jsx(_createMdxContent,{...props})}):_createMdxContent(props)}return{default:MDXContent};function _missingMdxReference(id,component){throw new Error(`Expected `+(component?`component`:`object`)+\" `\"+id+\"` to be defined: you likely forgot to import, pass, or provide it.\")}","frontmatter":{},"scope":{"config":{},"pageMetadata":{"title":"Run an Experiment to compare and evaluate responses from different models","sidebarTitle":"Compare and Evaluate Multiple Models","description":null,"href":"/guides/datasets/comparing-models"}}},"mdxSourceWithNoJs":null},"mdxExtracts":{"tableOfContents":[{"title":"Create or Import a Dataset","slug":"create-or-import-a-dataset","depth":3,"children":[]},{"title":"Configure Experiment","slug":"configure-experiment","depth":3,"children":[]},{"title":"Compare Responses","slug":"compare-responses","depth":3,"children":[]},{"title":"Run Evaluations","slug":"run-evaluations","depth":3,"children":[]},{"title":"View the Evaluation Metrics","slug":"view-the-evaluation-metrics","depth":3,"children":[]}],"codeExamples":{}},"jsLinks":[],"subdomain":"athinaai","siteName":"Athina","title":"Run an Experiment to compare and evaluate responses from different models - Athina","navMaps":{"firstHrefInVersion":{},"firstHrefInLanguage":{}},"originalFileLocation":"guides/datasets/comparing-models.mdx"},"headHtml":"\u003cscript data-mint-head>(function(){var d=document.documentElement,p=null;if(!p){try{p=localStorage.getItem(\"isDarkMode\")}catch(e){}if(p==='true')p='dark';else if(p==='false')p='light';if(p!=='light'\u0026\u0026p!=='dark'\u0026\u0026p!=='system')p=\"system\"}var r=p==='system'?(matchMedia(\"(prefers-color-scheme: dark)\").matches?'dark':'light'):p;d.classList.remove(r==='dark'?'light':'dark');d.classList.add(r);d.setAttribute(\"data-theme-preference\",p)})();\u003c/script>\u003cmeta data-mint-head name=\"application-name\" content=\"Athina\">\u003cmeta data-mint-head name=\"generator\" content=\"Mintlify\">\u003clink data-mint-head rel=\"apple-touch-icon\" type=\"image/png\" sizes=\"180x180\" href=\"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/_generated/favicon/apple-touch-icon.png?v=3\">\u003clink data-mint-head rel=\"icon\" type=\"image/png\" sizes=\"32x32\" href=\"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/_generated/favicon/favicon-32x32.png?v=3\">\u003clink data-mint-head rel=\"icon\" type=\"image/png\" sizes=\"16x16\" href=\"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/_generated/favicon/favicon-16x16.png?v=3\">\u003clink data-mint-head rel=\"shortcut icon\" type=\"image/x-icon\" href=\"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/_generated/favicon/favicon.ico?v=3\">\u003cmeta data-mint-head property=\"og:title\" content=\"Run an Experiment to compare and evaluate responses from different models - Athina\">\u003cmeta data-mint-head property=\"og:site_name\" content=\"Athina\">\u003cmeta data-mint-head property=\"og:url\" content=\"https://docs.athina.ai/guides/datasets/comparing-models\">\u003cmeta data-mint-head property=\"og:type\" content=\"website\">\u003cmeta data-mint-head property=\"og:image\" content=\"https://athinaai.mintlify.site/_mintlify/api/og/v2?division=Datasets\u0026amp;title=Run+an+Experiment+to+compare+and+evaluate+responses+from+different+models\u0026amp;theme=64652801f93e763e814fdf35\u0026amp;signature=7741c2a56972a2c4f046e9068c1591254f39c9a6019d44cb0e0fd3d5b0c22ecc\">\u003cmeta data-mint-head property=\"og:image:width\" content=\"1200\">\u003cmeta data-mint-head property=\"og:image:height\" content=\"630\">\u003cmeta data-mint-head name=\"twitter:card\" content=\"summary_large_image\">\u003cmeta data-mint-head name=\"twitter:title\" content=\"Run an Experiment to compare and evaluate responses from different models - Athina\">\u003cmeta data-mint-head name=\"twitter:image\" content=\"https://athinaai.mintlify.site/_mintlify/api/og/v2?division=Datasets\u0026amp;title=Run+an+Experiment+to+compare+and+evaluate+responses+from+different+models\u0026amp;theme=64652801f93e763e814fdf35\u0026amp;signature=7741c2a56972a2c4f046e9068c1591254f39c9a6019d44cb0e0fd3d5b0c22ecc\">\u003cmeta data-mint-head name=\"twitter:image:width\" content=\"1200\">\u003cmeta data-mint-head name=\"twitter:image:height\" content=\"630\">\u003cmeta data-mint-head name=\"msapplication-config\" content=\"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/_generated/favicon/browserconfig.xml?v=3\">\u003cmeta data-mint-head name=\"apple-mobile-web-app-title\" content=\"Athina\">\u003cmeta data-mint-head name=\"msapplication-TileColor\" content=\"#1976D2\">\u003clink data-mint-head rel=\"canonical\" href=\"https://docs.athina.ai/guides/datasets/comparing-models\">\u003clink data-mint-head rel=\"alternate\" type=\"application/xml\" href=\"/sitemap.xml\">\u003clink data-mint-head rel=\"alternate\" type=\"text/markdown\" href=\"/guides/datasets/comparing-models.md\">\u003cscript data-mint-head type=\"application/ld+json\">{\"@context\":\"https://schema.org\",\"@graph\":[{\"@type\":\"Organization\",\"@id\":\"https://docs.athina.ai/#organization\",\"name\":\"Athina\",\"url\":\"https://docs.athina.ai\",\"logo\":{\"@type\":\"ImageObject\",\"url\":\"https://mintlify.s3.us-west-1.amazonaws.com/athinaai/logo/light.svg\"}},{\"@type\":\"WebSite\",\"@id\":\"https://docs.athina.ai#website\",\"name\":\"Athina\",\"url\":\"https://docs.athina.ai\",\"publisher\":{\"@id\":\"https://docs.athina.ai/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https://docs.athina.ai/guides/datasets/comparing-models#webpage\",\"url\":\"https://docs.athina.ai/guides/datasets/comparing-models\",\"name\":\"Run an Experiment to compare and evaluate responses from different models\",\"isPartOf\":{\"@id\":\"https://docs.athina.ai#website\"},\"breadcrumb\":{\"@id\":\"https://docs.athina.ai/guides/datasets/comparing-models#breadcrumb\"}},{\"@type\":\"BreadcrumbList\",\"@id\":\"https://docs.athina.ai/guides/datasets/comparing-models#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Datasets\",\"item\":\"https://docs.athina.ai/guides/datasets/aws-bedrock-model\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Run an Experiment to compare and evaluate responses from different models\",\"item\":\"https://docs.athina.ai/guides/datasets/comparing-models\"}]},{\"@type\":[\"Article\",\"TechArticle\"],\"@id\":\"https://docs.athina.ai/guides/datasets/comparing-models#article\",\"headline\":\"Run an Experiment to compare and evaluate responses from different models\",\"name\":\"Run an Experiment to compare and evaluate responses from different models\",\"url\":\"https://docs.athina.ai/guides/datasets/comparing-models\",\"mainEntityOfPage\":{\"@id\":\"https://docs.athina.ai/guides/datasets/comparing-models#webpage\"},\"publisher\":{\"@id\":\"https://docs.athina.ai/#organization\"},\"isPartOf\":{\"@id\":\"https://docs.athina.ai#website\"}}]}\u003c/script>\u003cscript data-mint-head type=\"application/ld+json\">{\"@context\":\"https://schema.org\",\"@type\":\"WebSite\",\"name\":\"Athina\",\"publisher\":{\"@type\":\"Organization\",\"name\":\"Athina\"},\"provider\":{\"@type\":\"Organization\",\"name\":\"Mintlify\",\"url\":\"https://mintlify.com\"}}\u003c/script>","agentPointerHref":"/guides/datasets/comparing-models.md","siteName":"Athina","nav":{"global":null,"anchors":[{"anchor":"Documentation","icon":"book-open","tabs":[{"tab":"Docs","pages":[{"title":"Docs","href":"/overview"}]},{"tab":"API / SDK Reference","pages":[{"title":"API / SDK Reference","href":"/api-reference/logging/overview"}]},{"tab":"Guides","groups":[{"group":"Getting Started","pages":[{"title":"Getting Started Guides","sidebarTitle":"Overview","href":"/guides/overview"}]},{"group":"Prompts","pages":[{"title":"Prompt Comparison","sidebarTitle":"Prompt Comparison","icon":"grip-lines","href":"/guides/prompts/prompt-comparison"},{"title":"Prompt Versioning","sidebarTitle":"Prompt Versioning","icon":"code-compare","href":"/guides/prompts/prompt-versioning"}]},{"group":"Datasets","pages":[{"title":"Integrate AWS Bedrock Models","sidebarTitle":"AWS Bedrock Models","icon":"aws","href":"/guides/datasets/aws-bedrock-model"},{"title":"Preparing Data for Fine-Tuning","sidebarTitle":"Preparing Data for Fine-Tuning","icon":"table","href":"/guides/datasets/preparing-data-for-fine-tuning"},{"title":"Get Data from S3 Bucket","sidebarTitle":"Get Data from S3 Bucket","icon":"aws","description":"Step-by-step guide on retrieving S3 data in to Athina.","href":"/guides/datasets/get-data-from-s3-bucket"},{"title":"Run an Experiment to compare and evaluate responses from different models","sidebarTitle":"Compare and Evaluate Multiple Models","href":"/guides/datasets/comparing-models"},{"title":"Comparing datasets using Athina IDE","description":null,"href":"/guides/datasets/comparing-datasets"},{"title":"Prototype and Evaluate a Prompt Chain","href":"/guides/datasets/prototype-and-evaluate-prompt-chain"},{"title":"Run Prompts and Evaluate","href":"/guides/datasets/run-prompts-and-evaluate"}]},{"group":"Evals","pages":[{"title":"Pairwise Evaluation","sidebarTitle":"Pairwise Evaluation","icon":"scale-balanced","href":"/guides/evals/pairwise-evals"},{"title":"Evaluations in CI/CD Pipeline","sidebarTitle":"Evaluations in CI/CD Pipeline","icon":"infinity","href":"/guides/evals/evals-in-cicd-pipeline"},{"title":"Which evaluations to use for RAG applications?","sidebarTitle":"A guide to RAG evaluation","href":"/guides/evals/rag-eval-guide"},{"title":"Prompt Injection: Attacks and Defenses","href":"/guides/evals/prompt-injection"},{"title":"Running evals as real-time guardrails","href":"/guides/evals/running-evals-guardrails"},{"title":"Different stages of evaluation","sidebarTitle":"Different stages of evaluation","href":"/guides/evals/llm-eval-workflows"},{"title":"How can I improve the performance / reliability of my evals?","sidebarTitle":"Improving Eval performance","href":"/guides/evals/improving-eval-performance"},{"title":"How to Measure Retrieval Accuracy in RAG Applications Using Athina IDE","sidebarTitle":"Measuring Retrieval Accuracy in RAG Apps","href":"/guides/evals/measuring-retrieval-accuracy-in-rag"},{"title":"Pairwise Evaluation","icon":"columns-3","href":"/guides/evals/pairwise-evaluation"},{"title":"Evaluate Conversations","sidebarTitle":"Evaluate Conversations","icon":"comment","href":"/guides/evals/evaluate-conversations"},{"title":"RAG Evaluators","icon":"files","href":"/guides/evals/rag-evals"}]},{"group":"Experiments","pages":[{"title":"Compare Multiple Models","sidebarTitle":"Compare Multiple Models","icon":"head-side-gear","href":"/guides/experiments/compare-multiple-models"}]},{"group":"Flows","pages":[{"title":"Create and Share Flow","sidebarTitle":"Create and Share Flow","icon":"water","href":"/guides/flows/create-and-share-flow"}]}]},{"tab":"FAQs","pages":[{"title":"FAQs","href":"/faqs/data-policy"}]}]},{"anchor":"Open-Source Evals","href":"https://github.com/athina-ai/athina-evals","icon":"github"},{"anchor":"Blog","href":"https://blog.athina.ai/","icon":"book"},{"anchor":"Email us","href":"mailto:hello@athina.ai","icon":"envelope"},{"anchor":"Book a call","href":"https://cal.com/shiv-athina/30min","icon":"calendar"}]},"basePath":"","title":"Run an Experiment to compare and evaluate responses from different models - Athina","buildId":"dpl_DqYu5MPb49gWBBbUGgpHQUmCHY3P","deploymentId":"dpl_DqYu5MPb49gWBBbUGgpHQUmCHY3P","dynamics":[],"chunks":[]}