{"id":521098,"date":"2026-06-19T20:23:27","date_gmt":"2026-06-20T03:23:27","guid":{"rendered":"https:\/\/jorgep.com\/blog\/?p=521098"},"modified":"2026-06-19T20:39:50","modified_gmt":"2026-06-20T03:39:50","slug":"the-quest-for-token-efficiency-why-every-token-matters-now","status":"publish","type":"post","link":"https:\/\/jorgep.com\/blog\/the-quest-for-token-efficiency-why-every-token-matters-now\/","title":{"rendered":"The Quest for Token Efficiency: Why Every Token Matters Now"},"content":{"rendered":"\n<div class=\"wp-block-columns has-theme-palette-7-background-color has-background is-layout-flex wp-container-core-columns-is-layout-5dc627e1 wp-block-columns-is-layout-flex\" style=\"margin-top:0;margin-bottom:0\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:80%\">\n<p class=\"wp-block-paragraph\">Part of: <strong> <a href=\"https:\/\/jorgep.com\/blog\/series-ai-learnings\/\">AI Learning Series Here<\/a><\/strong><\/p>\n\n\n<style>.kadence-column395113_97b87a-23 > .kt-inside-inner-col,.kadence-column395113_97b87a-23 > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column395113_97b87a-23 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column395113_97b87a-23 > .kt-inside-inner-col{flex-direction:column;}.kadence-column395113_97b87a-23 > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column395113_97b87a-23 > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column395113_97b87a-23{position:relative;}@media all and (max-width: 1024px){.kadence-column395113_97b87a-23 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column395113_97b87a-23 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column395113_97b87a-23\"><div class=\"kt-inside-inner-col\"><style>.kadence-column510545_44b637-b5 > .kt-inside-inner-col{padding-top:var(--global-kb-spacing-xs, 1rem);padding-bottom:var(--global-kb-spacing-xs, 1rem);}.kadence-column510545_44b637-b5 > .kt-inside-inner-col,.kadence-column510545_44b637-b5 > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column510545_44b637-b5 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column510545_44b637-b5 > .kt-inside-inner-col{flex-direction:column;}.kadence-column510545_44b637-b5 > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column510545_44b637-b5 > .kt-inside-inner-col{background-color:var(--global-palette7, #EDF2F7);}.kadence-column510545_44b637-b5:hover > .kt-inside-inner-col{background-color:var(--global-palette8, #F7FAFC);background-image:none;}.kadence-column510545_44b637-b5 > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column510545_44b637-b5{position:relative;}@media all and (max-width: 1024px){.kadence-column510545_44b637-b5 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column510545_44b637-b5 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column510545_44b637-b5\"><div class=\"kt-inside-inner-col\"><style>.wp-block-kadence-advancedheading.kt-adv-heading510545_f7c5f8-ed, .wp-block-kadence-advancedheading.kt-adv-heading510545_f7c5f8-ed[data-kb-block=\"kb-adv-heading510545_f7c5f8-ed\"]{text-align:center;font-size:var(--global-kb-font-size-md, 1.25rem);font-style:normal;}.wp-block-kadence-advancedheading.kt-adv-heading510545_f7c5f8-ed mark.kt-highlight, .wp-block-kadence-advancedheading.kt-adv-heading510545_f7c5f8-ed[data-kb-block=\"kb-adv-heading510545_f7c5f8-ed\"] mark.kt-highlight{font-style:normal;color:#f76a0c;-webkit-box-decoration-break:clone;box-decoration-break:clone;padding-top:0px;padding-right:0px;padding-bottom:0px;padding-left:0px;}.wp-block-kadence-advancedheading.kt-adv-heading510545_f7c5f8-ed img.kb-inline-image, .wp-block-kadence-advancedheading.kt-adv-heading510545_f7c5f8-ed[data-kb-block=\"kb-adv-heading510545_f7c5f8-ed\"] img.kb-inline-image{width:150px;vertical-align:baseline;}<\/style>\n<p class=\"kt-adv-heading510545_f7c5f8-ed wp-block-kadence-advancedheading\" data-kb-block=\"kb-adv-heading510545_f7c5f8-ed\">Quick Links:&nbsp;<a href=\"https:\/\/jorgep.com\/blog\/resources-for-learning-ai\/\">Resources for Learning AI<\/a> | <a href=\"https:\/\/jorgep.com\/blog\/keeping-up-with-ai\/\">Keep up with AI<\/a> | <a href=\"https:\/\/jorgep.com\/blog\/list-of-ai-tools\/\" data-type=\"post\" data-id=\"402818\">List of AI Tools<\/a> | <a href=\"https:\/\/jorgep.com\/blog\/local-ai-series\/\" data-type=\"page\" data-id=\"519365\">Local AI<\/a> | <a href=\"https:\/\/jorgep.com\/blog\/tag\/ai-agents\/\" data-type=\"post_tag\" data-id=\"941\">AI Agents<\/a> |  <a href=\"https:\/\/jorgep.com\/blog\/work-beyond-tomorrow-series\/\" data-type=\"page\" data-id=\"365001\">Future of Work<\/a><\/p>\n<\/div><\/div>\n<\/div><\/div>\n\n\n<style>.kb-row-layout-id395113_d73e95-0d > .kt-row-column-wrap{align-content:start;}:where(.kb-row-layout-id395113_d73e95-0d > .kt-row-column-wrap) > .wp-block-kadence-column{justify-content:start;}.kb-row-layout-id395113_d73e95-0d > .kt-row-column-wrap{column-gap:var(--global-kb-gap-md, 2rem);row-gap:var(--global-kb-gap-none, 0rem );padding-top:var(--global-kb-spacing-xxs, 0.5rem);padding-bottom:var(--global-kb-spacing-xxs, 0.5rem);grid-template-columns:repeat(2, minmax(0, 1fr));}.kb-row-layout-id395113_d73e95-0d > .kt-row-layout-overlay{opacity:0.30;}@media all and (max-width: 1024px){.kb-row-layout-id395113_d73e95-0d > .kt-row-column-wrap{grid-template-columns:repeat(2, minmax(0, 1fr));}}@media all and (max-width: 767px){.kb-row-layout-id395113_d73e95-0d > .kt-row-column-wrap{grid-template-columns:minmax(0, 1fr);}}<\/style><div class=\"kb-row-layout-wrap kb-row-layout-id395113_d73e95-0d alignnone wp-block-kadence-rowlayout\"><div class=\"kt-row-column-wrap kt-has-2-columns kt-row-layout-equal kt-tab-layout-inherit kt-mobile-layout-row kt-row-valign-top\">\n<style>.kadence-column395113_df36f9-de > .kt-inside-inner-col,.kadence-column395113_df36f9-de > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column395113_df36f9-de > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column395113_df36f9-de > .kt-inside-inner-col{flex-direction:column;}.kadence-column395113_df36f9-de > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column395113_df36f9-de > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column395113_df36f9-de{position:relative;}@media all and (max-width: 1024px){.kadence-column395113_df36f9-de > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column395113_df36f9-de > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column395113_df36f9-de\"><div class=\"kt-inside-inner-col\"><style>.wp-block-kadence-advancedheading.kt-adv-heading395113_b3212c-b9, .wp-block-kadence-advancedheading.kt-adv-heading395113_b3212c-b9[data-kb-block=\"kb-adv-heading395113_b3212c-b9\"]{text-align:center;font-size:var(--global-kb-font-size-sm, 0.9rem);line-height:60px;font-style:normal;background-color:#f5a511;}.wp-block-kadence-advancedheading.kt-adv-heading395113_b3212c-b9 mark.kt-highlight, .wp-block-kadence-advancedheading.kt-adv-heading395113_b3212c-b9[data-kb-block=\"kb-adv-heading395113_b3212c-b9\"] mark.kt-highlight{font-style:normal;color:#f76a0c;-webkit-box-decoration-break:clone;box-decoration-break:clone;padding-top:0px;padding-right:0px;padding-bottom:0px;padding-left:0px;}.wp-block-kadence-advancedheading.kt-adv-heading395113_b3212c-b9 img.kb-inline-image, .wp-block-kadence-advancedheading.kt-adv-heading395113_b3212c-b9[data-kb-block=\"kb-adv-heading395113_b3212c-b9\"] img.kb-inline-image{width:150px;vertical-align:baseline;}<\/style>\n<p class=\"kt-adv-heading395113_b3212c-b9 wp-block-kadence-advancedheading\" data-kb-block=\"kb-adv-heading395113_b3212c-b9\">Subscribe to <a href=\"https:\/\/go.35s.be\/jtb\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>JorgeTechBits  newsletter<\/strong><\/a><\/p>\n<\/div><\/div>\n\n\n<style>.kadence-column395113_4b4b81-29 > .kt-inside-inner-col,.kadence-column395113_4b4b81-29 > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column395113_4b4b81-29 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column395113_4b4b81-29 > .kt-inside-inner-col{flex-direction:column;}.kadence-column395113_4b4b81-29 > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column395113_4b4b81-29 > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column395113_4b4b81-29{position:relative;}@media all and (max-width: 1024px){.kadence-column395113_4b4b81-29 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column395113_4b4b81-29 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column395113_4b4b81-29\"><div class=\"kt-inside-inner-col\"><\/div><\/div>\n\n<\/div><\/div><\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large is-resized\"><a href=\"htthttps:\/\/jorgep.com\/blog\/book-dont-just-chat-delegate\/\"><img loading=\"lazy\" decoding=\"async\" width=\"640\" height=\"1024\" src=\"https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01-640x1024.jpg\" alt=\"\" class=\"wp-image-520234\" style=\"aspect-ratio:0.6250142320391666;width:98px;height:auto\" srcset=\"https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01-640x1024.jpg 640w, https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01-188x300.jpg 188w, https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01-768x1229.jpg 768w, https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01-960x1536.jpg 960w, https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01-1280x2048.jpg 1280w, https:\/\/jorgep.com\/blog\/wp-content\/uploads\/CoverBook-01.jpg 1600w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption class=\"wp-element-caption\"><a href=\"https:\/\/jorgep.com\/blog\/book-series-ai-dont-just-chat\/\" data-type=\"page\" data-id=\"520242\">Check out the Book Series<\/a><\/figcaption><\/figure>\n<\/div><\/div>\n<\/div>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The artificial intelligence industry has experienced exponential growth in model capabilities over the past few years. As we have moved from models with billions of parameters to systems containing hundreds of billions of parameters, while expanding context windows into the millions of tokens, a new challenge has emerged: <strong>token efficiency<\/strong>. Every token carries a cost \u2014 consuming compute resources, increasing infrastructure demands, and directly impacting the economics and sustainability of AI applications. As organizations scale their use of AI, optimizing how models process, generate, and consume tokens has become a critical focus.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">At the same time, the conversation around reducing AI costs is increasingly shifting toward <strong>local AI<\/strong> \u2014 running models directly on devices or within a company\u2019s own infrastructure rather than relying exclusively on cloud-based AI services. In principle, this approach offers greater control, improved privacy, and potentially lower operating costs. However, there are important trade-offs to consider. While local AI can provide significant advantages for certain workloads, it does not necessarily mean companies can replicate the capabilities of today\u2019s leading frontier models. Understanding the balance between cost, performance, scalability, and control will be a key part of the next phase of AI adoption.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">First time I talked about <a href=\"https:\/\/jorgep.com\/blog\/tag\/tokens\/\" data-type=\"post_tag\" data-id=\"1017\">Token Economics<\/a> was back in 2024 in my blog:  <a href=\"https:\/\/jorgep.com\/blog\/understanding-ai-tokens-the-building-blocks-of-ai-applications\/\" data-type=\"post\" data-id=\"518307\">Understanding AI Tokens: The Building Blocks of AI Applications<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Why Token Efficiency Matters Now<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Several converging forces have made token efficiency impossible to ignore.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Cost Scaling<\/strong>: AI pricing is predominantly per-token. As applications process longer documents, maintain extended conversation histories, and generate elaborate outputs, costs scale linearly with token volume. A chatbot that handles ten-turn conversations on thousand-token contexts costs fundamentally more than one optimized for five-turn conversations on hundred-token inputs. At scale, these differences determine business model viability.<\/li>\n\n\n\n<li><strong>Latency and User Experience<\/strong>: Token generation happens sequentially. Longer outputs take more time to produce. Higher latency degrades user experience, particularly for interactive applications. Users tolerate milliseconds, not seconds, between query and response. Every unnecessary token adds perceptible delay.<\/li>\n\n\n\n<li><strong>Context Window Pressures<\/strong>: Modern models accept enormous contexts, from hundreds of thousands to millions of tokens. This capacity enables impressive capabilities \u2014 analyzing entire codebases, processing long documents, maintaining extended conversations. But filling those windows is expensive. Sending a million tokens to a frontier model can cost more than many users expect to pay for an entire month of service.<\/li>\n\n\n\n<li><strong>Environmental Considerations<\/strong>: Inference consumes energy. Longer contexts and larger outputs require more computation, which translates directly to greater electricity consumption and carbon emissions. Organizations with sustainability commitments must account for the environmental cost of every token.<\/li>\n\n\n\n<li><strong>Accessibility and Equity<\/strong>: High per-token costs limit who can build and use AI applications. Developers in regions with weaker currencies, students, hobbyists, and small organizations face barriers when token-heavy architectures become standard. Efficiency democratizes access.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">The Myth of Cheap AI: Why Local Models Are Not the Whole Story<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">There is a growing conversation that the answer to rising AI costs \u2014 especially token costs from using frontier models \u2014 is to bring AI models in-house and run them locally. On the surface, this makes a lot of sense: companies gain more control over their data, reduce dependency on external APIs, and can potentially lower long-term inference costs. However, the reality is more nuanced. Running models locally does not mean a company can simply download and host the same frontier models powering the leading AI platforms.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The key distinction is between <strong>frontier models<\/strong> and <strong>open-source\/open-weight models<\/strong>. Models such as GPT-class frontier systems or Claude-class frontier systems are generally not available for companies to deploy fully inside their own infrastructure. Local deployments typically rely on open-weight models such as Llama, Qwen, Mistral, or similar alternatives. These models can deliver excellent performance for many enterprise workloads, but they are not identical to the largest frontier systems. The future will likely be a hybrid approach: using smaller, efficient local models for high-volume tasks while reserving frontier models for complex reasoning and specialized workloads where maximum capability is worth the additional cost.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n<style>.kb-row-layout-id521098_5e9813-d2 > .kt-row-column-wrap{align-content:start;}:where(.kb-row-layout-id521098_5e9813-d2 > .kt-row-column-wrap) > .wp-block-kadence-column{justify-content:start;}.kb-row-layout-id521098_5e9813-d2 > .kt-row-column-wrap{column-gap:var(--global-kb-gap-md, 2rem);row-gap:var(--global-kb-gap-md, 2rem);padding-top:var(--global-kb-spacing-sm, 1.5rem);padding-bottom:var(--global-kb-spacing-sm, 1.5rem);grid-template-columns:repeat(2, minmax(0, 1fr));}.kb-row-layout-id521098_5e9813-d2 > .kt-row-layout-overlay{opacity:0.30;}@media all and (max-width: 1024px){.kb-row-layout-id521098_5e9813-d2 > .kt-row-column-wrap{grid-template-columns:repeat(2, minmax(0, 1fr));}}@media all and (max-width: 767px){.kb-row-layout-id521098_5e9813-d2 > .kt-row-column-wrap{grid-template-columns:minmax(0, 1fr);}}<\/style><div class=\"kb-row-layout-wrap kb-row-layout-id521098_5e9813-d2 alignnone wp-block-kadence-rowlayout\"><div class=\"kt-row-column-wrap kt-has-2-columns kt-row-layout-equal kt-tab-layout-inherit kt-mobile-layout-row kt-row-valign-top\">\n<style>.kadence-column521098_0ef05e-08 > .kt-inside-inner-col,.kadence-column521098_0ef05e-08 > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column521098_0ef05e-08 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column521098_0ef05e-08 > .kt-inside-inner-col{flex-direction:column;}.kadence-column521098_0ef05e-08 > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column521098_0ef05e-08 > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column521098_0ef05e-08{position:relative;}@media all and (max-width: 1024px){.kadence-column521098_0ef05e-08 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column521098_0ef05e-08 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column521098_0ef05e-08\"><div class=\"kt-inside-inner-col\">\n<h3 class=\"wp-block-heading\">Pros of bringing AI models in-house<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>1. Lower long-term token costs<\/strong><br>Once the infrastructure is in place, companies can avoid paying per-token API fees and run high-volume workloads at a predictable cost.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>2. Greater data control and privacy<\/strong><br>Sensitive information can remain inside company-controlled infrastructure, which can help with compliance, security requirements, and proprietary data protection.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>3. Customization and specialization<\/strong><br>Open models can be fine-tuned, optimized, and integrated deeply into internal workflows, allowing companies to build AI systems tailored to their specific business needs.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>4. Reduced vendor dependency<\/strong><br>Companies have more control over their AI stack and are less exposed to pricing changes, API limits, or platform changes from external providers.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>5. Better economics for specific workloads<\/strong><br>For repetitive tasks such as document processing, internal search, classification, summarization, and workflow automation, smaller local models can often deliver strong results at a lower cost.<\/p>\n<\/div><\/div>\n\n\n<style>.kadence-column521098_8909eb-81 > .kt-inside-inner-col,.kadence-column521098_8909eb-81 > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column521098_8909eb-81 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column521098_8909eb-81 > .kt-inside-inner-col{flex-direction:column;}.kadence-column521098_8909eb-81 > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column521098_8909eb-81 > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column521098_8909eb-81{position:relative;}@media all and (max-width: 1024px){.kadence-column521098_8909eb-81 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column521098_8909eb-81 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column521098_8909eb-81\"><div class=\"kt-inside-inner-col\">\n<h3 class=\"wp-block-heading\">Cons of bringing AI models in-house<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>1. You are usually not running the true frontier models<\/strong><br>The biggest limitation is that companies generally cannot simply self-host the same models used by leading AI providers. Local deployments usually rely on open-weight alternatives that may be less capable for complex reasoning, coding, or advanced tasks.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>2. Infrastructure costs can be significant<\/strong><br>Running AI locally requires GPUs, servers, storage, networking, monitoring, security, and specialized engineering expertise. The hardware investment can outweigh API costs for smaller workloads.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>3. Operational complexity<\/strong><br>Managing models is not just downloading a file and running it. Companies need to handle scaling, updates, performance tuning, security, model evaluation, and reliability.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>4. Faster model improvements from vendors<\/strong><br>Frontier AI providers are continuously improving their models. A self-hosted model can become outdated unless the company invests continuously in upgrades and testing.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>5. Talent requirements<\/strong><br>Building and operating an internal AI platform requires machine learning, infrastructure, and DevOps expertise that many organizations do not have internally.<\/p>\n<\/div><\/div>\n\n<\/div><\/div>\n\n\n<p class=\"wp-block-paragraph\">The practical reality is that most companies will not replace frontier AI services completely. A more realistic strategy is a <strong>hybrid AI architecture<\/strong>: local open models for high-volume, cost-sensitive workloads, combined with frontier models for the tasks where the highest intelligence and accuracy provide real business value.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Leveraging Local AI for Token Savings<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The rise of Local AI presents a promising avenue for reducing token costs while maintaining robust AI capabilities. By deploying open-source and small-scale models on local infrastructure, organizations can significantly cut down on token expenses associated with cloud-based AI services. Open-source models like LLaMA, Mistral, and DeepSeek offer flexibility and customization, allowing businesses to tailor them to specific needs without incurring per-token fees. Additionally, small models such as quantized versions of larger models or distilled variants can efficiently handle tasks like classification, summarization, and simple Q&amp;A, further reducing token consumption.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n<style>.kb-row-layout-id521098_6e16d1-5f > .kt-row-column-wrap{align-content:start;}:where(.kb-row-layout-id521098_6e16d1-5f > .kt-row-column-wrap) > .wp-block-kadence-column{justify-content:start;}.kb-row-layout-id521098_6e16d1-5f > .kt-row-column-wrap{column-gap:var(--global-kb-gap-md, 2rem);row-gap:var(--global-kb-gap-md, 2rem);padding-top:var(--global-kb-spacing-sm, 1.5rem);padding-bottom:var(--global-kb-spacing-sm, 1.5rem);grid-template-columns:repeat(2, minmax(0, 1fr));}.kb-row-layout-id521098_6e16d1-5f > .kt-row-layout-overlay{opacity:0.30;}@media all and (max-width: 1024px){.kb-row-layout-id521098_6e16d1-5f > .kt-row-column-wrap{grid-template-columns:repeat(2, minmax(0, 1fr));}}@media all and (max-width: 767px){.kb-row-layout-id521098_6e16d1-5f > .kt-row-column-wrap{grid-template-columns:minmax(0, 1fr);}}<\/style><div class=\"kb-row-layout-wrap kb-row-layout-id521098_6e16d1-5f alignnone wp-block-kadence-rowlayout\"><div class=\"kt-row-column-wrap kt-has-2-columns kt-row-layout-equal kt-tab-layout-inherit kt-mobile-layout-row kt-row-valign-top\">\n<style>.kadence-column521098_f7cc4b-be > .kt-inside-inner-col,.kadence-column521098_f7cc4b-be > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column521098_f7cc4b-be > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column521098_f7cc4b-be > .kt-inside-inner-col{flex-direction:column;}.kadence-column521098_f7cc4b-be > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column521098_f7cc4b-be > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column521098_f7cc4b-be{position:relative;}@media all and (max-width: 1024px){.kadence-column521098_f7cc4b-be > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column521098_f7cc4b-be > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column521098_f7cc4b-be\"><div class=\"kt-inside-inner-col\">\n<p class=\"wp-block-paragraph\"><strong>Pros:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Cost Efficiency:<\/strong>\u00a0Eliminates per-token fees, making it ideal for high-volume, repetitive tasks.<\/li>\n\n\n\n<li><strong>Data Privacy:<\/strong>\u00a0Local processing ensures sensitive data never leaves the organization, addressing compliance and security concerns.<\/li>\n\n\n\n<li><strong>Customization:<\/strong>\u00a0Models can be fine-tuned on proprietary data, improving performance for specific applications.<\/li>\n\n\n\n<li><strong>Latency Reduction:<\/strong>\u00a0On-premise deployment reduces network latency, leading to faster response times.<\/li>\n<\/ul>\n<\/div><\/div>\n\n\n<style>.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col,.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col:before{border-top-left-radius:0px;border-top-right-radius:0px;border-bottom-right-radius:0px;border-bottom-left-radius:0px;}.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm, 1rem);}.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col{flex-direction:column;}.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col > .aligncenter{width:100%;}.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col:before{opacity:0.3;}.kadence-column521098_c64f1b-f2{position:relative;}@media all and (max-width: 1024px){.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}@media all and (max-width: 767px){.kadence-column521098_c64f1b-f2 > .kt-inside-inner-col{flex-direction:column;justify-content:center;}}<\/style>\n<div class=\"wp-block-kadence-column kadence-column521098_c64f1b-f2\"><div class=\"kt-inside-inner-col\">\n<p class=\"wp-block-paragraph\"><strong>Cons:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Infrastructure Costs:<\/strong>\u00a0Requires investment in GPUs, storage, and maintenance, which may not be feasible for smaller organizations.<\/li>\n\n\n\n<li><strong>Expertise Required:<\/strong>\u00a0Setting up and managing local AI systems demands skilled personnel.<\/li>\n\n\n\n<li><strong>Limited Scalability:<\/strong>\u00a0Scaling local AI solutions can be challenging compared to cloud-based alternatives.<\/li>\n\n\n\n<li><strong>Model Performance:<\/strong>\u00a0Small models may not match the accuracy or capability of frontier models for complex tasks.<\/li>\n<\/ul>\n<\/div><\/div>\n\n<\/div><\/div>\n\n\n<p class=\"wp-block-paragraph\">By strategically integrating Local AI into their workflows, organizations can balance cost savings with performance, using local models for routine tasks while reserving cloud-based frontier models for more complex, high-stakes operations. A practical approach is to start small by identifying specific, repetitive tasks where local models can deliver comparable results to cloud solutions, thereby immediately cutting costs. Gradually, as infrastructure and expertise grow, organizations can expand the scope of local models to encompass more complex tasks, ensuring a scalable and cost-effective AI strategy.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Understanding Where Tokens Go<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Before optimizing, one must understand token consumption patterns. Tokens are spent in several distinct categories:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Category<\/th><th>Description<\/th><\/tr><\/thead><tbody><tr><td>System Prompts and Instructions<\/td><td>Base instructions that define model behavior, repeated with every request.<\/td><\/tr><tr><td>Context and History<\/td><td>Previous conversation turns or documents, typically dominating token usage.<\/td><\/tr><tr><td>Retrial-Augmented Generation<\/td><td>Retrieval systems that append documents, potentially adding thousands of tokens.<\/td><\/tr><tr><td>Multimodal Content<\/td><td>Tokens consumed by non-text data such as images, audio, and video.<\/td><\/tr><tr><td>Output Generation<\/td><td>Tokens consumed by verbose outputs and structured formats like JSON.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Strategies for Token Efficiency<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The quest for efficiency operates across multiple layers, from architecture choices to prompt engineering to runtime optimization.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Right-Size Your Model<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Choosing the correct model size is crucial. Not every task requires the largest available model. Simple classification, entity extraction, and short-form generation often perform adequately with smaller, faster models. Routing simple queries to efficient models while reserving frontier models for complex reasoning can reduce costs by orders of magnitude. Larger models aren&#8217;t always necessary. Here&#8217;s a comparison:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Model Type<\/th><th>Use Case<\/th><th>Cost\/Token<\/th><th>Performance<\/th><\/tr><\/thead><tbody><tr><td><strong>Smaller Models<\/strong><\/td><td>Simple tasks, short responses<\/td><td>Low<\/td><td>Adequate for basic tasks<\/td><\/tr><tr><td><strong>Large\/Frontier Models<\/strong><\/td><td>Complex reasoning tasks<\/td><td>High<\/td><td>Best for advanced tasks<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The economics are stark. A task that costs cents on a large model might cost fractions of a cent on a smaller alternative. At volume, this difference compounds into substantial savings.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Model Hosting Options<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Companies can host models internally in various ways, each with distinct characteristics:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Hosting Type<\/th><th>Example Models<\/th><th>Description<\/th><th>Benefits<\/th><\/tr><\/thead><tbody><tr><td><strong>API-Only Models<\/strong><\/td><td>OpenAI GPT models, Anthropic Claude<\/td><td>Run exclusively on vendor infrastructure.<\/td><td>Best performance, no infrastructure needed.<\/td><\/tr><tr><td><strong>Vendor-Hosted in Cloud<\/strong><\/td><td>Azure OpenAI, Amazon Bedrock, Google Vertex AI<\/td><td>Models deployed within the company&#8217;s cloud account, managed by the vendor.<\/td><td>Enhanced data privacy within cloud, vendor-managed scaling.<\/td><\/tr><tr><td><strong>Fully Self-Hosted Models<\/strong><\/td><td>Meta LLaMA, Mistral, DeepSeek, Alibaba Qwen<\/td><td>Hosted on the company&#8217;s own GPU clusters or private data centers.<\/td><td>Maximum control, data sovereignty, customization options.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">Token Efficiency and Cost Management<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">To reduce token costs, companies can strategically select models based on their requirements:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Model Category<\/th><th>Examples<\/th><th>Cost-Effectiveness<\/th><th>Typical Use Cases<\/th><\/tr><\/thead><tbody><tr><td><strong>Self-hosted Open-Weight Models<\/strong><\/td><td>Qwen, LLaMA, DeepSeek<\/td><td>Lowest marginal cost post-deployment<\/td><td>High-volume internal applications, compliance-sensitive tasks<\/td><\/tr><tr><td><strong>Small Commercial Models<\/strong><\/td><td>Claude Haiku, GPT Mini, Gemini Flash<\/td><td>Cost-effective for medium-scale operations<\/td><td>Mid-scale applications requiring moderate performance<\/td><\/tr><tr><td><strong>Frontier Flagship Models<\/strong><\/td><td>Claude Opus, GPT Flagship, Gemini Ultra<\/td><td>Premium cost, superior capabilities<\/td><td>Complex reasoning tasks, high-stakes applications<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">Detailed Insights on Self-Hostable Open-Weight Models<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Model Name<\/th><th>Developer<\/th><th>Strengths<\/th><th>Typical Applications<\/th><\/tr><\/thead><tbody><tr><td><strong>Meta LLaMA<\/strong><\/td><td>Meta<\/td><td>Versatile, strong community support<\/td><td>General AI tasks, research<\/td><\/tr><tr><td><strong>Mistral<\/strong><\/td><td>Mistral AI<\/td><td>High efficiency, good for European languages<\/td><td>Multilingual applications, enterprise solutions<\/td><\/tr><tr><td><strong>DeepSeek<\/strong><\/td><td>DeepSeek AI<\/td><td>Strong coding and math capabilities<\/td><td>Technical document processing, code generation<\/td><\/tr><tr><td><strong>Alibaba Qwen<\/strong><\/td><td>Alibaba<\/td><td>Strong multilingual capabilities, especially in Asian languages<\/td><td>Global applications, multilingual chatbots<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">Self-Hosting Considerations<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Self-hosting requires significant infrastructure and expertise. Here&#8217;s a breakdown:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Consideration<\/th><th>Details<\/th><\/tr><\/thead><tbody><tr><td><strong>Infrastructure<\/strong><\/td><td>Requires investment in GPUs and networking equipment.<\/td><\/tr><tr><td><strong>Expertise<\/strong><\/td><td>Needs skilled personnel for deployment, maintenance, and optimization.<\/td><\/tr><tr><td><strong>Customization<\/strong><\/td><td>Ability to fine-tune models on proprietary data for enhanced performance.<\/td><\/tr><tr><td><strong>Compliance<\/strong><\/td><td>Better alignment with regulatory requirements regarding data sovereignty.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Measuring and Monitoring<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Effective optimization requires measurement. Organizations should track:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Tokens per Request<\/strong>: Average input and output tokens per API call, broken down by endpoint and use case.<\/li>\n\n\n\n<li><strong>Cost per Task<\/strong>: End-to-end cost of completing user tasks, including all model calls, retrievals, and post-processing.<\/li>\n\n\n\n<li><strong>Latency Breakdown<\/strong>: Time spent on context preparation, model inference, and output processing.<\/li>\n\n\n\n<li><strong>Quality Metrics<\/strong>: Task success rates, user satisfaction, error rates. Efficiency must not come at the expense of effectiveness.<\/li>\n\n\n\n<li><strong>Provider Comparison<\/strong>: Cost and quality of equivalent tasks across different models and providers.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">This data enables continuous optimization and informs architectural decisions.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Real-World Application: Hybrid Approaches<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Many Fortune 500 companies are adopting hybrid approaches to balance performance and cost:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Routine Tasks:<\/strong> Use self-hosted models like Qwen or LLaMA for efficiency.<\/li>\n\n\n\n<li><strong>Complex Tasks:<\/strong> Escalate to frontier models such as GPT or Claude for quality.<\/li>\n\n\n\n<li><strong>Result:<\/strong> Significant cost reductions while maintaining high-quality outcomes for critical tasks.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">The Business Case for Efficiency<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Token efficiency is not merely an engineering optimization. It is a strategic business capability that directly correlates with improved economics and sustainability. <\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Competitive Pricing:<\/strong> Applications with efficient token usage can offer lower prices, higher margins, or more generous free tiers. They can serve more users on the same infrastructure. They can respond faster, improving user satisfaction and engagement.<\/li>\n\n\n\n<li><strong>Enhanced User Experience:<\/strong> Low latency leads to better engagement and user satisfaction.<\/li>\n\n\n\n<li><strong>Scalable Solutions:<\/strong> Applications that ignore token efficiency face cost structures that limit growth. Each new user adds linear cost. Scaling becomes expensive. Competition from more efficient alternatives erodes market position.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">The organizations that master token efficiency will define the economics of the AI application era. Those that treat tokens as unlimited will find their business models strained as usage scales.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The Future of Token Efficiency<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The industry is moving toward greater efficiency through multiple paths:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Mixture of Experts<\/strong>: Architectures that activate only relevant parameters per query reduce computational cost while maintaining large model capabilities.<\/li>\n\n\n\n<li><strong>Long-Context Optimizations<\/strong>: New attention mechanisms and state-space models process long contexts more efficiently than traditional transformers, reducing the cost of extended inputs.<\/li>\n\n\n\n<li><strong>Speculative Decoding<\/strong>: Draft models generate candidate tokens quickly, which the main model verifies. This parallelizes generation and reduces latency.<\/li>\n\n\n\n<li><strong>Hardware Specialization<\/strong>: Dedicated inference chips optimize for specific model architectures, delivering more tokens per watt than general-purpose GPUs.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Pricing Evolution<\/strong>: Providers may shift toward task-based pricing, quality-based pricing, or subscription models that decouple costs from raw token counts.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">My Bottom Line Thinking Right Now<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The future of AI adoption will not be defined by model size alone, but by how effectively organizations can balance intelligence, efficiency, cost, and control. Token efficiency is becoming a key factor in transforming powerful AI capabilities from impressive demonstrations into scalable, sustainable solutions that deliver real business value.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">As AI strategies mature, companies will need to make thoughtful decisions about where and how models are deployed \u2014 whether through frontier AI services, internally hosted open models, or hybrid approaches that combine both. The right choice will depend on the specific workload, security requirements, performance expectations, and economic considerations. Ultimately, success will come from using the right model, in the right environment, for the right task.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n<style>.wp-block-kadence-advancedheading.kt-adv-heading407818_afcbba-c7, .wp-block-kadence-advancedheading.kt-adv-heading407818_afcbba-c7[data-kb-block=\"kb-adv-heading407818_afcbba-c7\"]{font-size:var(--global-kb-font-size-sm, 0.9rem);font-style:normal;}.wp-block-kadence-advancedheading.kt-adv-heading407818_afcbba-c7 mark.kt-highlight, .wp-block-kadence-advancedheading.kt-adv-heading407818_afcbba-c7[data-kb-block=\"kb-adv-heading407818_afcbba-c7\"] mark.kt-highlight{font-style:normal;color:#f76a0c;-webkit-box-decoration-break:clone;box-decoration-break:clone;padding-top:0px;padding-right:0px;padding-bottom:0px;padding-left:0px;}.wp-block-kadence-advancedheading.kt-adv-heading407818_afcbba-c7 img.kb-inline-image, .wp-block-kadence-advancedheading.kt-adv-heading407818_afcbba-c7[data-kb-block=\"kb-adv-heading407818_afcbba-c7\"] img.kb-inline-image{width:150px;vertical-align:baseline;}<\/style>\n<p class=\"kt-adv-heading407818_afcbba-c7 wp-block-kadence-advancedheading\" data-kb-block=\"kb-adv-heading407818_afcbba-c7\">Have questions, ideas to share, or just want to connect? I\u2019d love to hear from you! Check out my <a href=\"https:\/\/jorgep.com\/blog\/about\/\">About Page<\/a> to learn more about me or connect with me.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>The artificial intelligence industry has experienced exponential growth in model capabilities over the past few years. As we have moved from models with billions of parameters to systems containing hundreds of billions of parameters, while expanding context windows into the millions of tokens, a new challenge has emerged: token efficiency. Every token carries a cost&#8230;<\/p>\n","protected":false},"author":2,"featured_media":521103,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_kad_blocks_custom_css":"","_kad_blocks_head_custom_js":"","_kad_blocks_body_custom_js":"","_kad_blocks_footer_custom_js":"","ngg_post_thumbnail":0,"episode_type":"","audio_file":"","podmotor_file_id":"","podmotor_episode_id":"","cover_image":"","cover_image_id":"","duration":"","filesize":"","filesize_raw":"","date_recorded":"","explicit":"","block":"","itunes_episode_number":"","itunes_title":"","itunes_season_number":"","itunes_episode_type":"","_kad_post_transparent":"","_kad_post_title":"","_kad_post_layout":"","_kad_post_sidebar_id":"","_kad_post_content_style":"","_kad_post_vertical_padding":"","_kad_post_feature":"","_kad_post_feature_position":"","_kad_post_header":false,"_kad_post_footer":false,"_kad_post_classname":"","footnotes":""},"categories":[441],"tags":[471,930,894,986,1017],"class_list":["post-521098","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tech-talk","tag-ai","tag-ai-series","tag-artificial-intelligence","tag-local-ai","tag-tokens"],"taxonomy_info":{"category":[{"value":441,"label":"Tech Talk"}],"post_tag":[{"value":471,"label":"AI"},{"value":930,"label":"AI Series"},{"value":894,"label":"artificial intelligence"},{"value":986,"label":"Local AI"},{"value":1017,"label":"Tokens"}]},"featured_image_src_large":["https:\/\/jorgep.com\/blog\/wp-content\/uploads\/FeaturedImage-TokenEcomonics.jpg",1000,375,false],"author_info":{"display_name":"Jorge Pereira","author_link":"https:\/\/jorgep.com\/blog\/author\/jorge\/"},"comment_info":0,"category_info":[{"term_id":441,"name":"Tech Talk","slug":"tech-talk","term_group":0,"term_taxonomy_id":451,"taxonomy":"category","description":"","parent":0,"count":729,"filter":"raw","cat_ID":441,"category_count":729,"category_description":"","cat_name":"Tech Talk","category_nicename":"tech-talk","category_parent":0}],"tag_info":[{"term_id":471,"name":"AI","slug":"ai","term_group":0,"term_taxonomy_id":481,"taxonomy":"post_tag","description":"","parent":0,"count":184,"filter":"raw"},{"term_id":930,"name":"AI Series","slug":"ai-series","term_group":0,"term_taxonomy_id":940,"taxonomy":"post_tag","description":"","parent":0,"count":192,"filter":"raw"},{"term_id":894,"name":"artificial intelligence","slug":"artificial-intelligence","term_group":0,"term_taxonomy_id":904,"taxonomy":"post_tag","description":"","parent":0,"count":186,"filter":"raw"},{"term_id":986,"name":"Local AI","slug":"local-ai","term_group":0,"term_taxonomy_id":996,"taxonomy":"post_tag","description":"","parent":0,"count":53,"filter":"raw"},{"term_id":1017,"name":"Tokens","slug":"tokens","term_group":0,"term_taxonomy_id":1027,"taxonomy":"post_tag","description":"","parent":0,"count":4,"filter":"raw"}],"_links":{"self":[{"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/posts\/521098","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/comments?post=521098"}],"version-history":[{"count":5,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/posts\/521098\/revisions"}],"predecessor-version":[{"id":521107,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/posts\/521098\/revisions\/521107"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/media\/521103"}],"wp:attachment":[{"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/media?parent=521098"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/categories?post=521098"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/jorgep.com\/blog\/wp-json\/wp\/v2\/tags?post=521098"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}