{"id":657,"date":"2026-05-20T20:10:12","date_gmt":"2026-05-20T20:10:12","guid":{"rendered":"https:\/\/convly.ai\/rtx-4090-vs-rtx-3090-for-ai\/"},"modified":"2026-06-10T05:05:10","modified_gmt":"2026-06-10T05:05:10","slug":"rtx-4090-vs-rtx-3090-for-ai","status":"publish","type":"post","link":"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/","title":{"rendered":"RTX 4090 versus RTX 3090 para IA em 2026: Vale a pena fazer a atualiza\u00e7\u00e3o?"},"content":{"rendered":"<p>For local AI work, the <strong>RTX 3090<\/strong> has aged into one of the best value cards ever made: 24 GB of VRAM on the used market for $700\u2013900. The <strong>RTX 4090<\/strong> doubles down \u2014 same 24 GB, but a far faster GPU at roughly $1,200\u20131,500 used in 2026.<\/p>\n<p>If both cards hold the same amount of memory, is the 4090 worth nearly double? The honest answer: <strong>it depends entirely on whether your time is the bottleneck.<\/strong><\/p>\n<div class=\"convly-tldr\">\n<h3>Principais conclus\u00f5es<\/h3>\n<ul>\n<li>Both cards have <strong>24 GB VRAM<\/strong> \u2014 they fit the exact same models. No model runs on one but not the other.<\/li>\n<li>The RTX 4090 is <strong>~1.7x faster<\/strong> for AI inference and <strong>~1.8x faster<\/strong> for fine-tuning.<\/li>\n<li>Para <strong>Stable Diffusion XL<\/strong>, expect ~18 it\/s on the 4090 vs ~10 it\/s on the 3090.<\/li>\n<li>The 3090 wins decisively on <strong>value-per-dollar<\/strong> and on dual-card builds (48 GB for ~$1,600).<\/li>\n<li>Buy the 4090 if iteration speed matters; buy the 3090 (or two) if VRAM capacity matters more than speed.<\/li>\n<\/ul>\n<\/div>\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_84 counter-flat ez-toc-counter ez-toc-container-direction\">\n<label for=\"ez-toc-cssicon-toggle-item-6a38af04bd95a\" class=\"ez-toc-cssicon-toggle-label\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Alternar<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #000000;color:#000000\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #000000;color:#000000\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewbox=\"0 0 24 24\" version=\"1.2\" baseprofile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/label><input type=\"checkbox\"  id=\"ez-toc-cssicon-toggle-item-6a38af04bd95a\"  aria-label=\"Alternar\" \/><nav><ul class='ez-toc-list ez-toc-list-level-1' ><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#At_a_glance\" >At a glance<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#VRAM_a_tie_that_changes_everything\" >VRAM: a tie that changes everything<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#Inference_benchmarks\" >Inference benchmarks<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#Fine-tuning_and_training\" >Fine-tuning and training<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#Power_and_heat\" >Power and heat<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#The_dual-3090_wildcard\" >The dual-3090 wildcard<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-7\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#Which_card_should_you_actually_buy\" >Which card should you actually buy?<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-8\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#FAQ\" >Perguntas frequentes<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-9\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#Verdict\" >Verdict<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-10\" href=\"https:\/\/convly.ai\/pt\/rtx-4090-vs-rtx-3090-for-ai\/#Related_articles\" >Artigos relacionados<\/a><\/li><\/ul><\/nav><\/div>\n<h2><span class=\"ez-toc-section\" id=\"At_a_glance\"><\/span>At a glance<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<table class=\"convly-vs\">\n<thead>\n<tr>\n<th>Especifica\u00e7\u00f5es<\/th>\n<th>RTX 4090<\/th>\n<th>RTX 3090<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Arquitetura<\/td>\n<td>Ada Lovelace AD102<\/td>\n<td>Ampere GA102<\/td>\n<\/tr>\n<tr>\n<td>N\u00facleos CUDA<\/td>\n<td class=\"convly-vs-winner\">16,384<\/td>\n<td>10,496<\/td>\n<\/tr>\n<tr>\n<td>VRAM<\/td>\n<td>24 GB GDDR6X<\/td>\n<td>24 GB GDDR6X<\/td>\n<\/tr>\n<tr>\n<td>Largura de banda de mem\u00f3ria<\/td>\n<td class=\"convly-vs-winner\">1,008 GB\/s<\/td>\n<td>936 GB\/s<\/td>\n<\/tr>\n<tr>\n<td>FP16 Tensor (dense)<\/td>\n<td class=\"convly-vs-winner\">~330 TFLOPS<\/td>\n<td>~142 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td>TDP<\/td>\n<td>450 W<\/td>\n<td class=\"convly-vs-winner\">350 W<\/td>\n<\/tr>\n<tr>\n<td>Launch price<\/td>\n<td>$1,599<\/td>\n<td class=\"convly-vs-winner\">$1,499<\/td>\n<\/tr>\n<tr>\n<td>Used price (2026)<\/td>\n<td>$1,200\u20131,500<\/td>\n<td class=\"convly-vs-winner\">$700\u2013900<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2><span class=\"ez-toc-section\" id=\"VRAM_a_tie_that_changes_everything\"><\/span>VRAM: a tie that changes everything<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The single most important number for local AI is VRAM, and here the two cards are identical: <strong>24 GB<\/strong>. That means any model that fits on one fits on the other:<\/p>\n<ul>\n<li><strong>Llama 3 8B<\/strong> e <strong>13B-class<\/strong> models run comfortably at full or near-full precision.<\/li>\n<li><strong>Llama 3 70B<\/strong> fits only at aggressive 4-bit quantization (Q4_K_M \u2248 40 GB) with partial CPU offload \u2014 painful on either card alone.<\/li>\n<li><strong>Stable Diffusion XL<\/strong> e <strong>Flux<\/strong> image models fit with room to spare.<\/li>\n<\/ul>\n<p>Because the memory ceiling is the same, the 4090 never unlocks a model the 3090 can&#8217;t touch. The 4090&#8217;s advantage is purely <strong>speed<\/strong> \u2014 it does the same work faster.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Inference_benchmarks\"><\/span>Inference benchmarks<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Para <strong>LLM inference<\/strong>, the gap tracks memory bandwidth and tensor throughput:<\/p>\n<table class=\"convly-vs\">\n<thead>\n<tr>\n<th>Workload<\/th>\n<th>RTX 4090<\/th>\n<th>RTX 3090<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Llama 3 8B Q4_K_M<\/td>\n<td class=\"convly-vs-winner\">~140 tok\/s<\/td>\n<td>~95 tok\/s<\/td>\n<\/tr>\n<tr>\n<td>Llama 3 13B-class Q4<\/td>\n<td class=\"convly-vs-winner\">~90 tok\/s<\/td>\n<td>~58 tok\/s<\/td>\n<\/tr>\n<tr>\n<td>SDXL 1024\u00d71024 (30 steps)<\/td>\n<td class=\"convly-vs-winner\">~18 it\/s<\/td>\n<td>~10 it\/s<\/td>\n<\/tr>\n<tr>\n<td>Flux.1 dev (1024px)<\/td>\n<td class=\"convly-vs-winner\">~2.4 s\/image<\/td>\n<td>~4.6 s\/image<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>The pattern is consistent: the 4090 lands around <strong>1.6\u20131.8x<\/strong> the 3090&#8217;s throughput. That is a real, felt difference \u2014 a Stable Diffusion batch that takes the 3090 ten minutes finishes in roughly six on the 4090.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Fine-tuning_and_training\"><\/span>Fine-tuning and training<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Para <strong>LoRA fine-tuning<\/strong> of a 7B\u20138B model, the 4090&#8217;s larger tensor-core throughput and faster FP16\/BF16 paths matter more than in inference. A typical LoRA run that takes the 3090 around five hours completes in roughly <strong>two-and-three-quarter hours<\/strong> on the 4090 \u2014 close to a 1.8x speedup.<\/p>\n<p>The 3090 has one quiet weakness here: it lacks the 4090&#8217;s improved FP8 support, so emerging FP8 training recipes either fall back to BF16 or don&#8217;t run at all. If you intend to follow cutting-edge training techniques, the 4090 ages better.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Power_and_heat\"><\/span>Power and heat<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The 3090 draws <strong>350 W<\/strong>; the 4090 draws <strong>450 W<\/strong> and can spike higher under sustained AI load. Over a year of heavy use that is a measurable difference on your power bill, and the 4090 demands a stronger PSU (850 W minimum, 1000 W recommended). The 3090 also runs hot on its GDDR6X memory modules \u2014 worth a thermal-pad replacement on used units.<\/p>\n<div class=\"convly-procons\">\n<div class=\"pros\">\n<h4>Choose the RTX 4090 if<\/h4>\n<ul>\n<li>You iterate constantly and value time over money<\/li>\n<li>You want FP8 support and better long-term software relevance<\/li>\n<li>You fine-tune models regularly, not just run inference<\/li>\n<\/ul>\n<\/div>\n<div class=\"cons\">\n<h4>Choose the RTX 3090 if<\/h4>\n<ul>\n<li>You want the most VRAM per dollar on the planet<\/li>\n<li>You plan a dual-card build (48 GB total for ~$1,600)<\/li>\n<li>Your workloads are batch jobs you can leave running overnight<\/li>\n<\/ul>\n<\/div>\n<\/div>\n<h2><span class=\"ez-toc-section\" id=\"The_dual-3090_wildcard\"><\/span>The dual-3090 wildcard<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Here is the argument that keeps the 3090 alive in 2026: <strong>two of them cost about the same as one used 4090<\/strong> and give you <strong>48 GB of pooled VRAM<\/strong>. With tensor parallelism (vLLM, ExLlamaV2), a dual-3090 rig runs <strong>Llama 3 70B<\/strong> entirely in VRAM \u2014 something no single consumer card except the RTX 5090 can do.<\/p>\n<p>You trade speed and power efficiency for capacity. For anyone whose real constraint is &#8220;I need to run bigger models,&#8221; two 3090s beat one 4090.<\/p>\n<p><!--ai-enriched--><\/p>\n<h2><span class=\"ez-toc-section\" id=\"Which_card_should_you_actually_buy\"><\/span>Which card should you actually buy?<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The specs only matter once you map them to a budget and a workload. Both cards carry the same 24&nbsp;GB of GDDR6X on a 384-bit bus, so the decision is rarely about whether a model <em>fits<\/em> &mdash; it is about how fast it runs and how much you are willing to pay for that speed. Use the tiers below as a starting point, then adjust for your local used-market prices.<\/p>\n<table class=\"convly-vs\">\n<tr>\n<th>Your situation<\/th>\n<th>Best pick<\/th>\n<th>Why<\/th>\n<\/tr>\n<tr>\n<td>Tightest budget, inference-only<\/td>\n<td><strong>Single used 3090<\/strong><\/td>\n<td>The best $\/VRAM on the consumer market for sub-70B models at Q4&ndash;Q5. Roughly half the price of a 4090 for the same memory ceiling.<\/td>\n<\/tr>\n<tr>\n<td>You fine-tune or train, or hate waiting<\/td>\n<td><strong>RTX 4090<\/strong><\/td>\n<td>Far higher compute and tensor throughput &mdash; roughly 1.5&ndash;2x in practice &mdash; meaningfully shortens training runs and diffusion batches. Time is the thing you are buying.<\/td>\n<\/tr>\n<tr>\n<td>You need 48&nbsp;GB for 70B at full context<\/td>\n<td><strong>Two 3090s<\/strong><\/td>\n<td>The 3090 is the last consumer card with NVLink, so it pools VRAM the 4090 cannot. (See the dual-3090 section above for the build caveats.)<\/td>\n<\/tr>\n<tr>\n<td>Latency-sensitive interactive use<\/td>\n<td><strong>RTX 4090<\/strong><\/td>\n<td>Higher tokens-per-second on a single card makes chat and agentic loops feel noticeably snappier.<\/td>\n<\/tr>\n<\/table>\n<p>A few honest caveats. A single 4090 will <strong>n\u00e3o<\/strong> run a model that a single 3090 cannot &mdash; they share the same 24&nbsp;GB wall, so do not pay the premium expecting to load bigger models on one card. If your only goal is to run a 70&nbsp;B model locally and you are patient, a used 3090 (or a pair of them) is the smarter money. If your time has a price &mdash; you iterate on LoRAs nightly, generate images in bulk, or run an agent that hammers the GPU all day &mdash; the 4090&#8217;s throughput pays for itself in hours saved.<\/p>\n<p>One non-obvious factor: <strong>condition and warranty<\/strong>. Most 3090s on the used market are several years old and many were run hard in mining or 24\/7 inference rigs. Budget for a possible thermal-pad replacement, inspect for sagging or fan wear, and favor sellers who offer returns. A 4090 bought used is younger and often still in warranty, which narrows the real-world price gap once you account for risk. Decide on the workload first, set a hard budget, and let those two numbers &mdash; not the spec sheet &mdash; choose the card.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"FAQ\"><\/span>Perguntas frequentes<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<h3>Is the RTX 4090 worth double the price of a 3090 for AI?<\/h3>\n<p>Only if speed is your bottleneck. The 4090 is ~1.7x faster but unlocks no new models, since both have 24 GB. If you run batch jobs overnight, the 3090&#8217;s value is unbeatable.<\/p>\n<h3>Can the RTX 3090 run Llama 3 70B?<\/h3>\n<p>Not comfortably on its own \u2014 70B at 4-bit needs ~40 GB. A single 3090 must offload layers to system RAM, which is slow. Two 3090s (48 GB pooled) run it well.<\/p>\n<h3>Which card is better for Stable Diffusion?<\/h3>\n<p>The RTX 4090, clearly \u2014 around 18 it\/s on SDXL versus 10 it\/s on the 3090. For image generation, where you iterate on prompts constantly, that speed gap is felt every minute.<\/p>\n<h3>Does the RTX 3090 still get good software support in 2026?<\/h3>\n<p>Yes. Ampere is fully supported by CUDA, PyTorch, vLLM, and llama.cpp. Its only gap is native FP8, which affects a small but growing set of training recipes.<\/p>\n<h3>How much does it cost to run these cards on electricity?<\/h3>\n<p>At the 2026 U.S. residential average of roughly $0.18\/kWh, a card pulling its full ~350&ndash;450&nbsp;W draws about 6&ndash;8&nbsp;cents per hour under load. Run an inference rig eight hours a day and that is only a few dollars a month &mdash; electricity is rarely the deciding factor for a single card. The 4090 is the more efficient of the two (more performance per watt), so it often costs <em>less<\/em> per task despite a higher peak draw. At idle both cards drop to a low draw &mdash; the 4090 dips to roughly 16&nbsp;W, the 3090 somewhat higher &mdash; so a machine left on but unused costs almost nothing.<\/p>\n<h3>Should I just buy an RTX 5090 instead?<\/h3>\n<p>Only if your budget stretches and you need the extra headroom. The 5090 brings 32&nbsp;GB of VRAM and a large generational compute jump, but in 2026 it still trades well above its $1,999 MSRP &mdash; commonly $2,500&ndash;$4,000+ &mdash; and supply remains tight. For models that fit in 24&nbsp;GB, a 3090 or 4090 delivers most of the practical value for a fraction of the outlay. The 5090 earns its price when you specifically need that 32&nbsp;GB ceiling or the fastest single-card inference available.<\/p>\n<h3>Which card holds its resale value better?<\/h3>\n<p>The 4090 has depreciated unusually slowly because RTX 5090 scarcity kept demand high, so it retains more of its value than a five-year-old card normally would. The 3090 is cheaper to buy and cheaper to exit, but its floor is propped up by one durable trait the newer GeForce cards dropped: NVLink. As long as builders want pooled-VRAM dual-card rigs, used 3090 demand &mdash; and resale value &mdash; stays resilient.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Verdict\"><\/span>Verdict<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>Both cards are excellent AI hardware in 2026. The <strong>RTX 4090<\/strong> is the better card in every raw metric and the right buy if you iterate fast and can absorb the price. The <strong>RTX 3090<\/strong> remains the value champion \u2014 and in a dual-card configuration it does something the 4090 simply cannot, running a 70B model fully in VRAM for less money. Match the card to your real constraint: speed, or capacity.<\/p>\n<p><!--related-block--><\/p>\n<div class=\"convly-related\">\n<h2><span class=\"ez-toc-section\" id=\"Related_articles\"><\/span>Artigos relacionados<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<ul>\n<li><a href=\"https:\/\/convly.ai\/pt\/rx-7900-xtx-vs-rtx-4090-for-ai\/\">AMD RX 7900 XTX versus RTX 4090 para IA em 2026: O ROCm consegue competir?<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/pt\/rtx-5080-vs-rtx-4080-super-for-ai\/\">RTX 5080 versus RTX 4080 Super para IA em 2026: Diferen\u00e7a geracional ou atualiza\u00e7\u00e3o lateral?<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/pt\/rtx-5070-ti-vs-rtx-4070-ti-super-for-ai\/\">RTX 5070 Ti versus RTX 4070 Ti Super para IA em 2026: Confronto na faixa intermedi\u00e1ria<\/a><\/li>\n<li><a href=\"https:\/\/convly.ai\/pt\/rtx-4060-ti-16gb-vs-rtx-3060-12gb-for-ai\/\">RTX 4060 Ti 16GB vs RTX 3060 12GB for AI in 2026: Best Budget GPU?<\/a><\/li>\n<\/ul>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Both the RTX 4090 and RTX 3090 ship 24 GB of VRAM \u2014 so for AI, the question isn&#8217;t whether a model fits, it&#8217;s how fast it runs. Here&#8217;s the benchmark-backed verdict.<\/p>","protected":false},"author":1,"featured_media":669,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[246],"tags":[281,256,352,280,282,353],"class_list":["post-657","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-comparisons","tag-ai-gpu","tag-local-llm","tag-rtx-3090","tag-rtx-4090","tag-stable-diffusion-benchmark","tag-used-gpu"],"_links":{"self":[{"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/posts\/657","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/comments?post=657"}],"version-history":[{"count":2,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/posts\/657\/revisions"}],"predecessor-version":[{"id":985,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/posts\/657\/revisions\/985"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/media\/669"}],"wp:attachment":[{"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/media?parent=657"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/categories?post=657"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/convly.ai\/pt\/wp-json\/wp\/v2\/tags?post=657"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}