{"id":407,"date":"2023-08-06T10:00:00","date_gmt":"2023-08-06T10:00:00","guid":{"rendered":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/"},"modified":"2023-08-06T10:00:00","modified_gmt":"2023-08-06T10:00:00","slug":"gpt-4-capacidades-reales","status":"publish","type":"post","link":"https:\/\/jacar.es\/en\/gpt-4-capacidades-reales\/","title":{"rendered":"GPT-4 In Depth: Real Capabilities vs Expectations"},"content":{"rendered":"<p>In March 2023, OpenAI launched GPT-4 with a presentation that promised \u201chuman-level performance on many benchmarks\u201d. Five months later, with thousands of real integrations, we\u2019re in a better position to judge: which capabilities held up, which were oversold, and where gaps remain versus alternatives like Claude 2 and LLaMA 2.<\/p>\n<h2 id=\"where-gpt-4-really-excels\">Where GPT-4 Really Excels<\/h2>\n<p>With the benchmark expanded to dozens of real cases, GPT-4 is consistently better at:<\/p>\n<ul>\n<li><strong>Complex chained reasoning<\/strong>. On problems requiring maintaining multiple variables, conditionals, and intermediate steps, GPT-4 hallucinates less and holds coherence better than any other currently available model.<\/li>\n<li><strong>Precise technical writing<\/strong>. Generating documentation, paper summaries, step-by-step explanations of complex concepts \u2014 especially in domains like programming, quantitative finance, or medicine \u2014 GPT-4 produces text requiring few editorial corrections.<\/li>\n<li><strong>Medium-complexity code<\/strong>. Not infallible, but on tasks like \u201crefactor this function\u201d, \u201cwrite tests for this component\u201d, or \u201cexplain what this legacy code does\u201d, it clearly beats GitHub Copilot when the IDE suggestion isn\u2019t enough.<\/li>\n<li><strong>Following very detailed instructions<\/strong>. A prompt with 15 specific constraints (\u201crespond in JSON with these keys, don\u2019t mention X, limit to 100 words\u201d) is followed much more faithfully than with GPT-3.5.<\/li>\n<\/ul>\n<h2 id=\"where-it-remains-frustrating\">Where It Remains Frustrating<\/h2>\n<p>But there are areas where GPT-4, despite the marketing, still systematically fails:<\/p>\n<h3 id=\"arithmetic\">Arithmetic<\/h3>\n<p>Surprisingly, GPT-4 makes trivial arithmetic errors relatively often. \u201cWhat is 2394 \u00d7 71?\u201d gives wrong results ~30% of the time. This improves dramatically with <a href=\"https:\/\/jacar.es\/code-interpreter-openai\/\">Code Interpreter<\/a>, which runs actual Python for calculations.<\/p>\n<h3 id=\"post-training-cutoff-information\">Post-training-cutoff information<\/h3>\n<p>The model was trained on data up to September 2021 (or April 2023 for updated versions). Questions about recent events, current library versions, or fresh news result in outdated or fabricated info. With the browsing plugin this improves, but latency and reliability drop.<\/p>\n<h3 id=\"consistency-across-conversations\">Consistency across conversations<\/h3>\n<p>The same question in two different conversations can get significantly different answers. For use cases requiring determinism (audits, reproducible validations), this forces more rigid prompting techniques or dropping <code>temperature<\/code> to 0 \u2014 which sometimes degrades quality.<\/p>\n<h3 id=\"large-numbers-and-very-long-contexts\">Large numbers and very long contexts<\/h3>\n<p>With 8k-32k tokens context (depending on version), GPT-4 stumbles when there\u2019s a lot of text. The famous <a href=\"https:\/\/arxiv.org\/abs\/2307.03172\">\u201clost in the middle\u201d<\/a> by Liu et al.\u00a0shows models ignore information placed in the middle of a long context. Claude 2 with 100k tokens does slightly better, but the problem exists in both.<\/p>\n<h2 id=\"gpt-4-vs-claude-2\">GPT-4 vs Claude 2<\/h2>\n<p><a href=\"https:\/\/www.anthropic.com\/news\/claude-2\">Claude 2<\/a>, released by Anthropic in July 2023, brings some notable advantages:<\/p>\n<ul>\n<li><strong>100k-token context window<\/strong>. Loads entire books, technical manuals, or long transcripts in a single prompt.<\/li>\n<li><strong>More conservative tone<\/strong>. Less prone to exaggeration or invention; when it doesn\u2019t know, it typically admits so.<\/li>\n<li><strong>Stricter built-in safety<\/strong>. For applications where minimising problematic responses matters, Claude 2 fails in fewer edge cases.<\/li>\n<\/ul>\n<p>Where GPT-4 wins: complex code, symbolic mathematics, multi-step reasoning on problems with more than 5 interacting entities.<\/p>\n<h2 id=\"gpt-4-vs-llama-2-70b\">GPT-4 vs LLaMA 2 70B<\/h2>\n<p>Compared with <a href=\"https:\/\/jacar.es\/llama-2-codigo-abierto\/\">LLaMA 2 70B<\/a>:<\/p>\n<ul>\n<li>GPT-4 clearly wins on complex reasoning and code.<\/li>\n<li>LLaMA 2 70B is competitive on summarisation, classification, simple Q&amp;A.<\/li>\n<li>LLaMA 2 has the absolute edge on privacy, cost at scale, and customisation.<\/li>\n<\/ul>\n<p>For any task where LLaMA 2 70B gives \u201cacceptable\u201d results, it\u2019s almost always the better choice \u2014 the quality delta rarely justifies the cost\/privacy tradeoff of GPT-4.<\/p>\n<h2 id=\"domain-evaluation\">Domain Evaluation<\/h2>\n<p>The only benchmark that matters is yours. A practical process for evaluating GPT-4 vs alternatives:<\/p>\n<ol type=\"1\">\n<li><strong>Select 20-30 representative prompts<\/strong> from your real application, with human-annotated \u201cideal\u201d answers.<\/li>\n<li><strong>Run each prompt through GPT-4, Claude 2, LLaMA 2<\/strong> and record responses.<\/li>\n<li><strong>Evaluate blindly<\/strong> (without knowing which model produced what): which came closest to the ideal?<\/li>\n<li><strong>Quantify total cost<\/strong>: per-token price \u00d7 expected volume + operational overhead.<\/li>\n<\/ol>\n<p>This process, covered in more detail in <a href=\"https:\/\/jacar.es\/prompt-engineering-maduro\/\">prompt engineering as a mature discipline<\/a>, often reveals surprises \u2014 sometimes Claude 2 wins where you expected GPT-4, or LLaMA 2 gives sufficient results at 1\/10 the cost.<\/p>\n<h2 id=\"responsible-use\">Responsible Use<\/h2>\n<p>A dimension worth remembering: GPT-4 generates plausible text even when wrong. For applications with real impact (medical, legal, financial decisions), model output must go through human validation or independent verification systems. The model has no way to know when it\u2019s confidently wrong, and that\u2019s dangerous in unsupervised flows.<\/p>\n<h2 id=\"conclusion\">Conclusion<\/h2>\n<p>GPT-4 is the most capable general-purpose model available in 2023, but \u201cmost capable\u201d doesn\u2019t mean \u201cbest choice for everything\u201d. Mature teams evaluate by use case, not by model reputation. In many scenarios, Claude 2 or LLaMA 2 offer better value\/cost ratios; in others, GPT-4 remains the unsurpassed standard. Team sophistication is measured in knowing which is which.<\/p>\n<p>Follow us on jacar.es for more on LLMs, AI evaluation, and product architecture.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Five months after GPT-4&#8217;s release, it&#8217;s time to separate real capabilities from hype. Where it excels and where it still fails.<\/p>\n","protected":false},"author":1,"featured_media":408,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[22,19],"tags":[129,130,77,56,76,51],"class_list":["post-407","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-inteligencia-artificial","category-tecnologia","tag-benchmarks","tag-evaluacion","tag-gpt-4","tag-ia-generativa","tag-llm","tag-openai"],"translation":{"provider":"WPGlobus","version":"3.0.2","language":"en","enabled_languages":["es","en"],"languages":{"es":{"title":true,"content":true,"excerpt":true},"en":{"title":true,"content":true,"excerpt":true}}},"gutentor_comment":0,"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.4 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>GPT-4 In Depth: Real Capabilities vs Expectations - Jacar<\/title>\n<meta name=\"description\" content=\"Honest GPT-4 analysis after 5 months in production: strengths in reasoning and code, weaknesses in math and currency, comparison with Claude 2 and LLaMA 2.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/jacar.es\/gpt-4-capacidades-reales\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"GPT-4 In Depth: Real Capabilities vs Expectations - Jacar\" \/>\n<meta property=\"og:description\" content=\"Honest GPT-4 analysis after 5 months in production: strengths in reasoning and code, weaknesses in math and currency, comparison with Claude 2 and LLaMA 2.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/jacar.es\/gpt-4-capacidades-reales\/\" \/>\n<meta property=\"og:site_name\" content=\"Jacar\" \/>\n<meta property=\"article:published_time\" content=\"2023-08-06T10:00:00+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\/wp-content\/uploads\/2020\/09\/favicon.png\" \/>\n\t<meta property=\"og:image:width\" content=\"252\" \/>\n\t<meta property=\"og:image:height\" content=\"229\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"javi\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"javi\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/\"},\"author\":{\"name\":\"javi\",\"@id\":\"https:\\\/\\\/jacar.es\\\/#\\\/schema\\\/person\\\/54a7f7b4224b38fafc9866eb3e614208\"},\"headline\":\"GPT-4 In Depth: Real Capabilities vs Expectations\",\"datePublished\":\"2023-08-06T10:00:00+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/\"},\"wordCount\":1611,\"publisher\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\\\/wp-content\\\/uploads\\\/2023\\\/08\\\/19205410\\\/jwp-1122798-11712.jpg\",\"keywords\":[\"benchmarks\",\"evaluacion\",\"gpt-4\",\"ia generativa\",\"llm\",\"openai\"],\"articleSection\":[\"Inteligencia Artificial\",\"Tecnolog\u00eda\"],\"inLanguage\":\"en-US\"},{\"@type\":[\"WebPage\",\"ItemPage\"],\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/\",\"url\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/\",\"name\":\"GPT-4 In Depth: Real Capabilities vs Expectations - Jacar\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\\\/wp-content\\\/uploads\\\/2023\\\/08\\\/19205410\\\/jwp-1122798-11712.jpg\",\"datePublished\":\"2023-08-06T10:00:00+00:00\",\"description\":\"Honest GPT-4 analysis after 5 months in production: strengths in reasoning and code, weaknesses in math and currency, comparison with Claude 2 and LLaMA 2.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#primaryimage\",\"url\":\"https:\\\/\\\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\\\/wp-content\\\/uploads\\\/2023\\\/08\\\/19205410\\\/jwp-1122798-11712.jpg\",\"contentUrl\":\"https:\\\/\\\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\\\/wp-content\\\/uploads\\\/2023\\\/08\\\/19205410\\\/jwp-1122798-11712.jpg\",\"width\":1200,\"height\":674,\"caption\":\"Interfaz digital representando inteligencia artificial\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/jacar.es\\\/gpt-4-capacidades-reales\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Portada\",\"item\":\"https:\\\/\\\/jacar.es\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"GPT-4 en profundidad: capacidades reales frente a expectativas\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/jacar.es\\\/#website\",\"url\":\"https:\\\/\\\/jacar.es\\\/\",\"name\":\"Jacar\",\"description\":\"Passion for Technology\",\"publisher\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/jacar.es\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/jacar.es\\\/#organization\",\"name\":\"Jacar\",\"url\":\"https:\\\/\\\/jacar.es\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/jacar.es\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/jacar.es\\\/wp-content\\\/uploads\\\/2020\\\/09\\\/favicon.png\",\"contentUrl\":\"https:\\\/\\\/jacar.es\\\/wp-content\\\/uploads\\\/2020\\\/09\\\/favicon.png\",\"width\":252,\"height\":229,\"caption\":\"Jacar\"},\"image\":{\"@id\":\"https:\\\/\\\/jacar.es\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/in\\\/javiercanetearroyo\\\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/jacar.es\\\/#\\\/schema\\\/person\\\/54a7f7b4224b38fafc9866eb3e614208\",\"name\":\"javi\",\"sameAs\":[\"https:\\\/\\\/jacar.es\"],\"url\":\"https:\\\/\\\/jacar.es\\\/en\\\/author\\\/javi\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"GPT-4 In Depth: Real Capabilities vs Expectations - Jacar","description":"Honest GPT-4 analysis after 5 months in production: strengths in reasoning and code, weaknesses in math and currency, comparison with Claude 2 and LLaMA 2.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/","og_locale":"en_US","og_type":"article","og_title":"GPT-4 In Depth: Real Capabilities vs Expectations - Jacar","og_description":"Honest GPT-4 analysis after 5 months in production: strengths in reasoning and code, weaknesses in math and currency, comparison with Claude 2 and LLaMA 2.","og_url":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/","og_site_name":"Jacar","article_published_time":"2023-08-06T10:00:00+00:00","og_image":[{"width":252,"height":229,"url":"https:\/\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\/wp-content\/uploads\/2020\/09\/favicon.png","type":"image\/png"}],"author":"javi","twitter_card":"summary_large_image","twitter_misc":{"Written by":"javi","Est. reading time":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#article","isPartOf":{"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/"},"author":{"name":"javi","@id":"https:\/\/jacar.es\/#\/schema\/person\/54a7f7b4224b38fafc9866eb3e614208"},"headline":"GPT-4 In Depth: Real Capabilities vs Expectations","datePublished":"2023-08-06T10:00:00+00:00","mainEntityOfPage":{"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/"},"wordCount":1611,"publisher":{"@id":"https:\/\/jacar.es\/#organization"},"image":{"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#primaryimage"},"thumbnailUrl":"https:\/\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\/wp-content\/uploads\/2023\/08\/19205410\/jwp-1122798-11712.jpg","keywords":["benchmarks","evaluacion","gpt-4","ia generativa","llm","openai"],"articleSection":["Inteligencia Artificial","Tecnolog\u00eda"],"inLanguage":"en-US"},{"@type":["WebPage","ItemPage"],"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/","url":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/","name":"GPT-4 In Depth: Real Capabilities vs Expectations - Jacar","isPartOf":{"@id":"https:\/\/jacar.es\/#website"},"primaryImageOfPage":{"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#primaryimage"},"image":{"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#primaryimage"},"thumbnailUrl":"https:\/\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\/wp-content\/uploads\/2023\/08\/19205410\/jwp-1122798-11712.jpg","datePublished":"2023-08-06T10:00:00+00:00","description":"Honest GPT-4 analysis after 5 months in production: strengths in reasoning and code, weaknesses in math and currency, comparison with Claude 2 and LLaMA 2.","breadcrumb":{"@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/jacar.es\/gpt-4-capacidades-reales\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#primaryimage","url":"https:\/\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\/wp-content\/uploads\/2023\/08\/19205410\/jwp-1122798-11712.jpg","contentUrl":"https:\/\/jcs-wp-jacar-es.fsn1.your-objectstorage.com\/wp-content\/uploads\/2023\/08\/19205410\/jwp-1122798-11712.jpg","width":1200,"height":674,"caption":"Interfaz digital representando inteligencia artificial"},{"@type":"BreadcrumbList","@id":"https:\/\/jacar.es\/gpt-4-capacidades-reales\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Portada","item":"https:\/\/jacar.es\/"},{"@type":"ListItem","position":2,"name":"GPT-4 en profundidad: capacidades reales frente a expectativas"}]},{"@type":"WebSite","@id":"https:\/\/jacar.es\/#website","url":"https:\/\/jacar.es\/","name":"Jacar","description":"Passion for Technology","publisher":{"@id":"https:\/\/jacar.es\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/jacar.es\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/jacar.es\/#organization","name":"Jacar","url":"https:\/\/jacar.es\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/jacar.es\/#\/schema\/logo\/image\/","url":"https:\/\/jacar.es\/wp-content\/uploads\/2020\/09\/favicon.png","contentUrl":"https:\/\/jacar.es\/wp-content\/uploads\/2020\/09\/favicon.png","width":252,"height":229,"caption":"Jacar"},"image":{"@id":"https:\/\/jacar.es\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.linkedin.com\/in\/javiercanetearroyo\/"]},{"@type":"Person","@id":"https:\/\/jacar.es\/#\/schema\/person\/54a7f7b4224b38fafc9866eb3e614208","name":"javi","sameAs":["https:\/\/jacar.es"],"url":"https:\/\/jacar.es\/en\/author\/javi\/"}]}},"_links":{"self":[{"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/posts\/407","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/comments?post=407"}],"version-history":[{"count":0,"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/posts\/407\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/media\/408"}],"wp:attachment":[{"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/media?parent=407"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/categories?post=407"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/jacar.es\/en\/wp-json\/wp\/v2\/tags?post=407"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}