{"id":18837,"date":"2026-09-04T01:30:54","date_gmt":"2026-09-04T05:30:54","guid":{"rendered":"https:\/\/wp.glbgpt.com\/?p=18837"},"modified":"2026-09-04T01:30:55","modified_gmt":"2026-09-04T05:30:55","slug":"gemini-3-8-flash-review","status":"publish","type":"post","link":"https:\/\/wp.glbgpt.com\/id\/hub\/gemini-3-8-flash-review","title":{"rendered":"Ulasan Singkat Gemini 3.8: Harga, Pengujian, dan Kesimpulan"},"content":{"rendered":"<style>\n.gemini38-review{--ink:#172033;--muted:#5e6878;--line:#dce3ea;--soft:#f5f7fa;--blue:#1769aa;--blue-soft:#eaf4fb;--coral:#d45f4c;--green:#16765a;--yellow:#fff7df;max-width:980px;margin:0 auto;color:var(--ink);font:16px\/1.7 Arial,Helvetica,sans-serif}.gemini38-review h1,.gemini38-review h2,.gemini38-review h3{line-height:1.22;color:#101827;letter-spacing:0}.gemini38-review h1{font-size:3rem;margin:0 0 18px}.gemini38-review h2{font-size:1.85rem;margin:52px 0 16px;border-top:1px solid var(--line);padding-top:26px}.gemini38-review h3{font-size:1.22rem;margin:28px 0 10px}.gemini38-review p{margin:0 0 18px}.gemini38-review a{color:var(--blue);text-decoration:underline}.gemini38-review .eyebrow{color:var(--coral);font-size:.78rem;font-weight:700;text-transform:uppercase}.gemini38-review .lede{font-size:1.12rem;color:#344054}.gemini38-review .notice,.gemini38-review .figure{border:1px solid var(--line);border-radius:8px;background:#fff;padding:20px;margin:24px 0}.gemini38-review .notice{background:var(--yellow);border-color:#ead79a}.gemini38-review .grid{display:grid;grid-template-columns:repeat(4,1fr);gap:12px;margin:24px 0}.gemini38-review .stat{background:var(--soft);border:1px solid var(--line);border-top:4px solid var(--blue);border-radius:8px;padding:16px}.gemini38-review .stat:nth-child(2){border-top-color:var(--coral)}.gemini38-review .stat:nth-child(3){border-top-color:var(--green)}.gemini38-review .stat:nth-child(4){border-top-color:#c69214}.gemini38-review .stat strong{display:block;font-size:1.35rem}.gemini38-review .stat span{display:block;color:var(--muted);font-size:.88rem}.gemini38-review table{width:100%;border-collapse:collapse;margin:22px 0;font-size:.94rem}.gemini38-review th,.gemini38-review td{border:1px solid var(--line);padding:12px;text-align:left;vertical-align:top}.gemini38-review th{background:#eef3f7}.gemini38-review .task-card{border:1px solid #cfd8e3;border-radius:8px;margin:28px 0;overflow:hidden;background:#fff}.gemini38-review .task-head{background:#edf5fb;padding:17px 20px;border-bottom:1px solid #cfd8e3}.gemini38-review .task-head h3{margin:0}.gemini38-review .task-body{padding:20px}.gemini38-review .task-grid{display:grid;grid-template-columns:1fr 1fr;gap:16px}.gemini38-review .slot{background:var(--soft);border:1px solid #d7dfe8;border-radius:6px;padding:14px;margin-top:12px;min-width:0}.gemini38-review .slot h4{margin:0 0 7px;font-size:.94rem}.gemini38-review .metrics{color:#334155;font-size:.85rem;border-bottom:1px solid #d7dfe8;padding-bottom:9px}.gemini38-review .output{background:#fff;border-left:3px solid #8ab7d4;padding:10px 12px;margin:10px 0 0}.gemini38-review pre{white-space:pre-wrap;overflow-wrap:anywhere;background:#18212c;color:#eef5fb;border-radius:5px;padding:12px;font:13px\/1.55 Consolas,monospace;margin:10px 0}.gemini38-review .winner{background:var(--blue-soft);border-left:4px solid var(--blue);padding:14px 16px;margin-top:16px}.gemini38-review .winner.old{background:#eef8f3;border-left-color:var(--green)}.gemini38-review .inconclusive{background:var(--yellow);border-left-color:#c69214}.gemini38-review .toc{background:#f8fafc;border:1px solid var(--line);padding:18px 22px}.gemini38-review .toc ol{margin:0;padding-left:22px}.gemini38-review .bars{margin:18px 0}.gemini38-review .bar-row{display:grid;grid-template-columns:180px 1fr 95px;gap:12px;align-items:center;margin:12px 0}.gemini38-review .bar{height:15px;background:#dfe6ec;border-radius:3px;overflow:hidden}.gemini38-review .bar span{display:block;height:100%;background:var(--blue)}.gemini38-review .bar.coral span{background:var(--coral)}.gemini38-review .small{color:var(--muted);font-size:.9rem}@media(max-width:700px){.gemini38-review h1{font-size:2.2rem}.gemini38-review .grid,.gemini38-review .task-grid{grid-template-columns:1fr 1fr}.gemini38-review .bar-row{grid-template-columns:125px 1fr 75px}}@media(max-width:560px){.gemini38-review .grid,.gemini38-review .task-grid{grid-template-columns:1fr}.gemini38-review h2{font-size:1.5rem}}\n<\/style>\n\n\n\n<div class=\"gemini38-review\">\n<p class=\"eyebrow\">Model review \u00b7 Official benchmarks \u00b7 Same-prompt API tests<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p class=\"lede\">Gemini 3.8 Flash is Google&#8217;s new general-availability Flash model for long-horizon software engineering, agents, and complex knowledge work. The launch claims are strong. Our first controlled tests are much less tidy.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>This Gemini 3.8 Flash review separates official facts from hands-on evidence. Google&#8217;s pages establish the release date, model status, context window, pricing, tools, and benchmark claims. Seven same-prompt API tasks show what the model actually returned beside Gemini 3.7 Flash.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>If you want to repeat the comparison without managing separate model tabs, <a href=\"https:\/\/www.glbgpt.com\/home?inviter=hub_popup&amp;login=1\">GlobalGPT<\/a> provides one workspace for testing multiple AI models. For context on the previous generation, our <a href=\"https:\/\/www.glbgpt.com\/hub\/gemini-3-6-flash-review\/\">Gemini 3.6 Ulasan Singkat<\/a> shows how quickly the Flash line has been changing.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"notice\"><strong>Status bukti:<\/strong> six matched tasks produced usable answers from both models. One research-synthesis task ended mid-sentence for both and is excluded from the quality score. Artificial Analysis was unreachable during this review, so no third-party leaderboard number is presented as verified.<\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"grid\"><div class=\"stat\"><strong>1\u20134<\/strong><span>Task wins: 3.8 vs 3.7<\/span><\/div><div class=\"stat\"><strong>1<\/strong><span>Dasi berkualitas<\/span><\/div><div class=\"stat\"><strong>1<\/strong><span>Inconclusive task<\/span><\/div><div class=\"stat\"><strong>6\/6<\/strong><span>Completed pairs favored 3.7 on speed<\/span><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<nav class=\"toc\" aria-label=\"Daftar isi\"><strong>Dalam ulasan ini<\/strong><ol><li><a href=\"#short-answer\">Jawaban singkat<\/a><\/li><li><a href=\"#what-is\">What Gemini 3.8 Flash is<\/a><\/li><li><a href=\"#specs\">Specs and availability<\/a><\/li><li><a href=\"#pricing\">Harga<\/a><\/li><li><a href=\"#benchmarks\">Uji kinerja resmi<\/a><\/li><li><a href=\"#method\">Test method<\/a><\/li><li><a href=\"#real-tests\">Real-world tests<\/a><\/li><li><a href=\"#scorecard\">Scorecard<\/a><\/li><li><a href=\"#limitations\">Keterbatasan<\/a><\/li><li><a href=\"#verdict\">Keputusan akhir<\/a><\/li><li><a href=\"#faq\">PERTANYAAN YANG SERING DIAJUKAN<\/a><\/li><\/ol><\/nav>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"short-answer\">Gemini 3.8 Flash Review: The Short Answer<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p><strong>Gemini 3.8 Flash did not beat Gemini 3.7 Flash in this first seven-task review.<\/strong> Across the six usable pairs, 3.8 won the frontend planning task, tied on a Python bug fix, and lost four tasks on status precision, data reasoning, bounded agent planning, and SEO editing. The remaining research task was inconclusive.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The gap was not about raw knowledge. It came from small defects in finished work: one wrong denominator claim, looser status labels, less economical prose, and a less concrete tool plan. Those are exactly the details that affect whether an answer can be used immediately.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"figure\"><h3>Ringkasan ulasan<\/h3><table><tr><th>Ukuran<\/th><th>Gemini 3.8 Flash<\/th><th>Gemini 3,7 Flash<\/th><\/tr><tr><td>Kemenangan tugas<\/td><td>1<\/td><td><strong>4<\/strong><\/td><\/tr><tr><td>Ties<\/td><td>1<\/td><td>1<\/td><\/tr><tr><td>Completed-task elapsed time<\/td><td>49.266s<\/td><td><strong>42.909s<\/strong><\/td><\/tr><tr><td>Token keluaran tugas yang telah diselesaikan<\/td><td>9,199<\/td><td><strong>7,489<\/strong><\/td><\/tr><tr><td>Estimated completed-task cost<\/td><td>$0.036121<\/td><td><strong>$0.028569<\/strong><\/td><\/tr><\/table><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p class=\"small\">These results describe six small text tasks on one API route and one configuration. They are not a universal ranking of either model.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"what-is\">What Is Gemini 3.8 Flash?<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p><a href=\"https:\/\/deepmind.google\/models\/gemini\/flash\/\">Google describes Gemini 3.8 Flash<\/a> as its most intelligent Flash workhorse for coding and agents, with particular emphasis on long-horizon software engineering and knowledge work. It is based on Gemini 3.7 Flash rather than a completely separate architecture.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The positioning matters because this is not merely a cheaper general chatbot. Google is pitching 3.8 Flash as a production agent model that can reason across documents, tools, code, and longer task sequences while preserving Flash-level latency and scale.<\/p>\n<\/div>\n\n\n\n<figure class=\"wp-block-image size-large\" style=\"margin:24px auto\"><img decoding=\"async\" style=\"width:100%;height:auto;border:1px solid #dce3ea;border-radius:8px\" src=\"https:\/\/wp.glbgpt.com\/wp-content\/uploads\/2026\/09\/google-gemini-3-8-flash-official-overview.jpg\" alt=\"Google DeepMind official Gemini 3.8 Flash product page\"><figcaption class=\"wp-element-caption\">Google positions Gemini 3.8 Flash as its Flash model for complex agentic tasks at scale. Source: <a href=\"https:\/\/deepmind.google\/models\/gemini\/flash\/\">Google DeepMind<\/a>.<\/figcaption><\/figure>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Gemini 3.8 Flash Release Date<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The <a href=\"https:\/\/deepmind.google\/models\/model-cards\/gemini-3-8-flash\/\">kartu model resmi<\/a> was published on <strong>September 2, 2026<\/strong>. Google&#8217;s product page lists the model as generally available rather than preview-only.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>What Changed From Gemini 3.7 Flash?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The official emphasis shifts toward longer-running agent tasks and finished artifacts. Google highlights software engineering, rigorous reasoning, tool execution, multimodal understanding, and complex enterprise work. In our small test pack, that positioning translated into the strongest frontend plan, but not more reliable finished answers across the board.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"specs\">Gemini 3.8 Flash Specs and Availability<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<table><tr><th>Spesifikasi<\/th><th>Gemini 3.8 Flash<\/th><th>Bukti<\/th><\/tr><tr><td>ID model API<\/td><td><code>gemini-3.8-flash<\/code><\/td><td>Official pricing page and returned API responses<\/td><\/tr><tr><td>Status<\/td><td>General availability<\/td><td><a href=\"https:\/\/deepmind.google\/models\/gemini\/flash\/\">Google DeepMind product page<\/a><\/td><\/tr><tr><td>Konteks masukan<\/td><td>Hingga 1 juta token<\/td><td>Kartu model resmi<\/td><\/tr><tr><td>Jumlah teks maksimum yang dapat ditampilkan<\/td><td>64K token<\/td><td>Kartu model resmi<\/td><\/tr><tr><td>Inputs<\/td><td>Text, images, audio, video, and PDFs<\/td><td>Halaman produk resmi<\/td><\/tr><tr><td>Output<\/td><td>Teks<\/td><td>Kartu model resmi<\/td><\/tr><tr><td>Dukungan alat<\/td><td>Function calling, Search as a tool, computer use<\/td><td>Halaman produk resmi<\/td><\/tr><tr><td>Official access surfaces<\/td><td>Gemini App, Gemini Enterprise Agent Platform, Google AI Studio, Gemini API, AI Mode, and Google Antigravity<\/td><td>Kartu model resmi<\/td><\/tr><tr><td>Batas akhir pengetahuan<\/td><td>March 2026 for some domains; January 2025 for others<\/td><td>Kartu model resmi<\/td><\/tr><\/table>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>A 1M-token window is capacity, not proof of perfect long-document recall. The test pack below includes a short status-sensitive extraction task, but it does not establish million-token performance.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"pricing\">Gemini 3.8 Flash Pricing and API Cost<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The <a href=\"https:\/\/ai.google.dev\/gemini-api\/docs\/pricing\">Halaman resmi harga API Gemini<\/a> lists a launch rate through December 31, 2026 of <strong>$0,75 per 1 juta token masukan<\/strong> dan <strong>$3,75 per 1 juta token yang dihasilkan<\/strong>, including thinking tokens. On January 1, 2027, those rates rise to $1.50 and $7.50.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"figure\"><h3>Published price change<\/h3><div class=\"bars\"><div class=\"bar-row\"><span>Input \u00b7 2026<\/span><div class=\"bar\"><span style=\"width:50%\"><\/span><\/div><strong>$0.75<\/strong><\/div><div class=\"bar-row\"><span>Input \u00b7 2027<\/span><div class=\"bar coral\"><span style=\"width:100%\"><\/span><\/div><strong>$1.50<\/strong><\/div><div class=\"bar-row\"><span>Output \u00b7 2026<\/span><div class=\"bar\"><span style=\"width:50%\"><\/span><\/div><strong>$3.75<\/strong><\/div><div class=\"bar-row\"><span>Output \u00b7 2027<\/span><div class=\"bar coral\"><span style=\"width:100%\"><\/span><\/div><strong>$7.50<\/strong><\/div><\/div><p class=\"small\">Paid-tier price per 1M tokens. The January 2027 rates are double the introductory 2026 rates.<\/p><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Gemini 3.8 Flash and Gemini 3.7 Flash had the same published introductory rate in our test period. Even so, 3.8 cost more in every completed pair because its recorded token usage was higher. Across T02\u2013T07, the estimates were $0.036121 for 3.8 and $0.028569 for 3.7.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>For a different point on the speed-and-cost curve, see our <a href=\"https:\/\/www.glbgpt.com\/hub\/gemini-3-5-flash-lite-review\/\">Gemini 3.5 Flash Lite review<\/a>. A Lite model serves a different workload, so its lower price should not be treated as an automatic quality win.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"benchmarks\">Gemini 3.8 Flash Official Benchmarks<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Google&#8217;s performance page emphasizes four areas: DeepSWE v1.1, Vals Finance Agent v2, Harvey&#8217;s Legal Agent Benchmark, and HLE-Verified. Its charts show Gemini 3.8 Flash ahead of Gemini 3.7 Flash on all three evaluations with exact published percentages.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<table><tr><th>Evaluasi resmi<\/th><th>Gemini 3.8 Flash<\/th><th>Gemini 3,7 Flash<\/th><th>Batas bukti<\/th><\/tr><tr><td>DeepSWE v1.1<\/td><td>Above 70% in Google&#8217;s chart<\/td><td>Lower than 3.8 in the chart<\/td><td>Google says 3.8 leads in the efficient area; the public graphic does not expose a precise numeric label.<\/td><\/tr><tr><td>Vals Finance Agent v2<\/td><td><strong>61.4%<\/strong><\/td><td>59.0%<\/td><td>Official Google chart; the benchmark&#8217;s full methodology still matters.<\/td><\/tr><tr><td>Harvey&#8217;s Legal Agent Benchmark<\/td><td><strong>10.0%<\/strong><\/td><td>8.8%<\/td><td>This benchmark result is not a guarantee of legal accuracy in deployment.<\/td><\/tr><tr><td>HLE-Verified<\/td><td><strong>54.9%<\/strong><\/td><td>53.6%<\/td><td>It does not directly measure SEO editing, UI planning, or response economy.<\/td><\/tr><\/table>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<figure class=\"wp-block-image size-large\" style=\"margin:24px auto\"><img decoding=\"async\" style=\"width:100%;height:auto;border:1px solid #dce3ea;border-radius:8px\" src=\"..\/assets\/google-gemini-3-8-flash-deepswe-benchmark.png\" alt=\"Google DeepMind DeepSWE v1.1 benchmark chart for Gemini 3.8 Flash\"><figcaption class=\"wp-element-caption\">Google&#8217;s official DeepSWE v1.1 chart places Gemini 3.8 Flash above 70% and in the chart&#8217;s most-efficient area. Source: <a href=\"https:\/\/deepmind.google\/models\/gemini\/flash\/#performance\">Google DeepMind performance page<\/a>.<\/figcaption><\/figure>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"notice\"><strong>Third-party benchmark note:<\/strong> Artificial Analysis could not be reached during research on September 4, 2026. This draft therefore uses official claims with clear attribution and leaves the independent leaderboard screenshot pending.<\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"method\">How We Tested Gemini 3.8 Flash<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Both models received the same task prompt, temperature of 0, a 3,200-token output ceiling, and the same API route. T02 was rerun once for both models after the first 3.8 request failed technically; only the matched rerun is scored.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<table><tr><th>Lapangan yang terekam<\/th><th>How it was used<\/th><\/tr><tr><td>Returned model ID<\/td><td>Confirmed that the requested model answered.<\/td><\/tr><tr><td>Real output<\/td><td>Compared correctness, completeness, status fidelity, and editing required.<\/td><\/tr><tr><td>Jumlah detik yang telah berlalu<\/td><td>Measured local end-to-end response time, not provider-side latency.<\/td><\/tr><tr><td>Token masukan dan keluaran<\/td><td>Supported task-level cost estimates at the official introductory rate.<\/td><\/tr><tr><td>Finish reason<\/td><td>Separated usable answers from incomplete or budget-limited output.<\/td><\/tr><\/table>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"real-tests\">Gemini 3.8 Flash Real-World Test Results<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Each card keeps the task, the actual returned work, the measurements, and the editorial verdict together. The excerpts are shortened for readability but preserve the deciding details.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 1: Research Synthesis<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Produce a 180-word buying recommendation from a locked fact set, with three reasons, two unknowns, and a test plan.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">18.211s \u00b7 361 input \/ 3,196 output tokens \u00b7 est. $0.012256<\/p><p class=\"output\">\u201cWe recommend selecting Model A&#8230; 3. Cost Parity: The introductory API price is the same for Model A and Model B\u201d<\/p><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">13.490s \u00b7 108 input \/ 3,196 output tokens \u00b7 est. $0.012066<\/p><p class=\"output\">\u201cWe directly recommend piloting and purchasing Model A&#8230; The introductory API price remains the exact same for both models through December 3\u201d<\/p><\/div><\/div><p>Both answers stop mid-sentence before the required unknowns and test plan. The 3.8 response reports <code>berhenti<\/code>, while 3.7 reports <code>panjang<\/code>; neither finish reason changes the usability judgment.<\/p><div class=\"winner inconclusive\"><strong>Verdict: Inconclusive.<\/strong> Neither model completed the requested deliverable.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 2: Python Bug Fix<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Return the three highest-value paid records, preserve input order on ties, handle empty input, explain the root cause, and add two tests.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">7.969s \u00b7 344 input \/ 1,517 output tokens \u00b7 est. $0.005947<\/p><pre><code>paid.sort(key=lambda r: r[\"amount\"], reverse=True)\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\nreturn paid[:3]<\/code><\/pre><p class=\"output\">It explained Python&#8217;s stable sort and tested ranking, ties, filtering, and empty input.<\/p><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">7.818s \u00b7 91 input \/ 1,359 output tokens \u00b7 est. $0.005164<\/p><pre><code>paid.sort(key=lambda r: r[\"amount\"], reverse=True)\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\nreturn paid[:3]<\/code><\/pre><p class=\"output\">It returned the same minimal fix and covered the same acceptance criteria.<\/p><\/div><\/div><p>The code is functionally equivalent. The 3.7 answer is slightly shorter and cheaper, but that is not enough to separate output quality.<\/p><div class=\"winner\"><strong>Verdict: Tie.<\/strong> Both produced a correct, usable patch.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 3: Textual Frontend Implementation Plan<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Plan five prioritized fixes, responsive behavior, and QA for a described checkout. No screenshot was supplied.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">10.128s \u00b7 340 input \/ 1,509 output tokens \u00b7 est. $0.005914<\/p><p class=\"output\">It specified a 58\/42 desktop grid, 48\u201352px CTA, nearby trust note, contextual discount alert, mobile order-summary accordion, focus states, and overflow checks.<\/p><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">8.149s \u00b7 87 input \/ 1,416 output tokens \u00b7 est. $0.005375<\/p><p class=\"output\">It proposed a similar grid and mobile stack, but introduced \u201c256-bit Encrypted Checkout,\u201d a detail absent from the task.<\/p><\/div><\/div><p>Both plans are implementable. The 3.8 plan has the stronger QA checklist and avoids inventing a specific encryption claim.<\/p><div class=\"winner\"><strong>Verdict: Gemini 3.8 Flash.<\/strong> More complete and safer to hand to a frontend team.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 4: Status-Aware Extraction<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Summarize Orion and preserve confirmed, planned, approved, pending, unknown, and unverified states in a fact table.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">6.292s \u00b7 371 input \/ 1,122 output tokens \u00b7 est. $0.004486<\/p><p class=\"output\">It labeled the second pilot \u201cConfirmed \/ Pending\u201d and wrote that the first pilot \u201csuccessfully handled\u201d 82% of cases.<\/p><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">7.126s \u00b7 118 input \/ 1,152 output tokens \u00b7 est. $0.004409<\/p><p class=\"output\">It kept the rollout \u201cPlanned (Unapproved),\u201d legal review \u201cPlanned (Pending),\u201d and production readiness \u201cUnknown.\u201d<\/p><\/div><\/div><p>Both retain the core facts, but 3.7 separates approval, planning, pending review, and unknown status more precisely. Both add \u201csuccessfully\u201d to the neutral 82% pilot statement.<\/p><div class=\"winner old\"><strong>Verdict: Gemini 3.7 Flash, slight win.<\/strong> Ketepatan status yang lebih baik.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 5: Data Reasoning<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Calculate ROAS and paid conversion rate, handle a zero-spend denominator correctly, rank channels, and recommend a conditional $300 split.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">11.074s \u00b7 377 input \/ 2,355 output tokens \u00b7 est. $0.009114<\/p><p class=\"output\">It correctly ranked Email 7.0\u00d7, Search 3.0\u00d7, Social 2.6\u00d7, and excluded Partner ROAS. It then said zero spend creates a zero denominator for both ROAS and CPA.<\/p><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">9.106s \u00b7 124 input \/ 1,969 output tokens \u00b7 est. $0.007477<\/p><p class=\"output\">It returned the same correct rankings and treated Partner ROAS as not measurable, then proposed $200 Email \/ $100 Search when email can scale.<\/p><\/div><\/div><p>The 3.8 answer makes a real mathematical error: CPA is spend divided by conversions, so $0 divided by 4 is $0, not a zero-denominator case.<\/p><div class=\"winner old\"><strong>Verdict: Gemini 3.7 Flash.<\/strong> Correct reasoning without the extra denominator mistake.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 6: Bounded Agentic Planning<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Use four named tools to collect one official API price, stop on unofficial or ambiguous evidence, and define completion in under 220 words.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">5.603s \u00b7 354 input \/ 948 output tokens \u00b7 est. $0.003821<\/p><p class=\"output\">It mapped search, fetch, parse, and persist steps, then added authenticity, availability, and price-clarity gates.<\/p><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">5.127s \u00b7 101 input \/ 894 output tokens \u00b7 est. $0.003428<\/p><p class=\"output\">It used explicit tool-call syntax, named <code>model_pricing.csv<\/code>, required distinct input and output rates, and defined one verified row as completion.<\/p><\/div><\/div><p>Both plans respect the stop conditions. The 3.7 plan is more concrete about the available calls and the saved artifact without pretending the tools ran.<\/p><div class=\"winner old\"><strong>Verdict: Gemini 3.7 Flash, slight win.<\/strong> More execution-ready at lower response cost.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"task-card\"><div class=\"task-head\"><h3>Test 7: SEO Writing and Editing<\/h3><\/div><div class=\"task-body\"><p><strong>Tugas:<\/strong> Return HTML only, use \u201cGemini 3.8 Flash review\u201d exactly once, preserve four locked facts, and avoid unsupported results, prices, dates, or access claims.<\/p><div class=\"task-grid\"><div class=\"slot\"><h4>Gemini 3.8 Flash<\/h4><p class=\"metrics\">8.200s \u00b7 379 input \/ 1,748 output tokens \u00b7 est. $0.006839<\/p><pre><code>&lt;p&gt;Our Gemini 3.8 Flash review evaluates the Google model by comparing official benchmark context with same-prompt API tests, with hands-on results pending as the evaluation records:&lt;\/p&gt;<\/code><\/pre><\/div><div class=\"slot\"><h4>Gemini 3,7 Flash<\/h4><p class=\"metrics\">5.583s \u00b7 126 input \/ 699 output tokens \u00b7 est. $0.002716<\/p><pre><code>&lt;p&gt;This Gemini 3.8 Flash review examines the new Google model by comparing official benchmark context with same-prompt API tests, with hands-on results currently pending.&lt;\/p&gt;<\/code><\/pre><\/div><\/div><p>Both satisfy the structure and locked facts. The 3.7 sentence is cleaner and reads like publishable SEO copy; 3.8&#8217;s \u201cas the evaluation records\u201d construction is awkward.<\/p><div class=\"winner old\"><strong>Verdict: Gemini 3.7 Flash.<\/strong> Better editorial finish with fewer tokens.<\/div><\/div><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"scorecard\">Gemini 3.8 Flash Test Scorecard<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<div class=\"figure\"><table><tr><th>Tugas<\/th><th>Pemenang<\/th><th>Deciding difference<\/th><\/tr><tr><td>Sintesis penelitian<\/td><td>Tidak pasti<\/td><td>Both incomplete<\/td><\/tr><tr><td>Python bug fix<\/td><td>Dasi<\/td><td>Equivalent code and coverage<\/td><\/tr><tr><td>Frontend plan<\/td><td><strong>Gemini 3.8 Flash<\/strong><\/td><td>Better QA and no invented encryption claim<\/td><\/tr><tr><td>Status extraction<\/td><td>Gemini 3,7 Flash<\/td><td>More precise state labels<\/td><\/tr><tr><td>Penalaran data<\/td><td>Gemini 3,7 Flash<\/td><td>No CPA denominator error<\/td><\/tr><tr><td>Perencanaan agen<\/td><td>Gemini 3,7 Flash<\/td><td>More concrete calls and artifact<\/td><\/tr><tr><td>Penyuntingan SEO<\/td><td>Gemini 3,7 Flash<\/td><td>Cleaner publishable sentence<\/td><\/tr><\/table><\/div>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The result is a useful counterweight to the launch story. Gemini 3.8 Flash may be stronger on the official agent and long-horizon evaluations Google chose to highlight, yet 3.7 produced better finished work in four of our small controlled tasks.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"limitations\">Gemini 3.8 Flash Limitations We Observed<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The biggest issue was not a dramatic failure. It was <strong>extra correction work<\/strong>: a denominator mistake, vague status labeling, an awkward SEO sentence, and more tokens used on every completed task. T01 also ended before completing a tightly specified 180-word deliverable.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>This test pack used text only. It does not evaluate image, audio, video, PDFs, computer use, true long-context retrieval, or a multi-hour agent run. Google&#8217;s model card also notes possible hallucinations, occasional slowness or timeouts, and higher token use at stronger effort levels.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"verdict\">Gemini 3.8 Flash Review: Final Verdict<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p><strong>Gemini 3.8 Flash is a meaningful official upgrade for agentic and long-horizon workloads, but it is not an automatic upgrade for every everyday task.<\/strong> It won one of six usable pairs in this review, while Gemini 3.7 Flash won four and used less time, fewer output tokens, and lower estimated cost.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The practical lesson is to test the work you actually ship. If your workload is long-running software engineering or document-heavy agent execution, Google&#8217;s official benchmark direction makes 3.8 worth a serious trial. If your priority is concise editing, small data analyses, or status-sensitive summaries, our first results say 3.7 remains highly competitive.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The same pattern appeared in our earlier <a href=\"https:\/\/www.glbgpt.com\/hub\/gemini-3-6-flash-vs-3-1-pro\/\">Perbandingan Gemini 3.6 Flash vs Gemini 3.1 Pro<\/a>: generation labels do not replace side-by-side output inspection. Run a recurring prompt in <a href=\"https:\/\/www.glbgpt.com\/home?inviter=hub_popup&amp;login=1\">GlobalGPT<\/a>, inspect the finished answer, and choose from evidence rather than the version number.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h2 id=\"faq\">Gemini 3.8 Flash FAQ<\/h2>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>When was Gemini 3.8 Flash released?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Google DeepMind published the Gemini 3.8 Flash model card on September 2, 2026.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Is Gemini 3.8 Flash generally available?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Yes. Google&#8217;s official product page lists Gemini 3.8 Flash as generally available.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>What is the Gemini 3.8 Flash API model ID?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>The official model ID is <code>gemini-3.8-flash<\/code>, which also matched the returned model ID in our API tests.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>How much does Gemini 3.8 Flash cost?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Through December 31, 2026, Google lists $0.75 per 1M input tokens and $3.75 per 1M output tokens, including thinking tokens. Starting January 1, 2027, the rates rise to $1.50 and $7.50.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Is Gemini 3.8 Flash free?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Google&#8217;s pricing page lists a free tier with limited model access and free input and output tokens. Limits and model availability can differ by account and route.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>What is the Gemini 3.8 Flash context window?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Gemini 3.8 Flash supports up to 1M input tokens and up to 64K tokens of text output, according to the official model card.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Does Gemini 3.8 Flash support images, audio, video, and PDFs?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Yes as inputs. Google&#8217;s product page lists text, image, video, audio, and PDF inputs; the model card lists text output.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Does Gemini 3.8 Flash support tools?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Google lists function calling, Search as a tool, and computer use. Tool availability can depend on the product and API route.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Is Gemini 3.8 Flash better than Gemini 3.7 Flash?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Not in every task. Gemini 3.8 Flash won one of six usable task pairs in this review, Gemini 3.7 Flash won four, and one pair tied. Google&#8217;s official evaluations emphasize stronger long-horizon agent and software-engineering performance.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Is Gemini 3.8 Flash good for coding?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Both models produced the correct minimal Python fix in our coding task. Google&#8217;s official positioning and DeepSWE claim make 3.8 especially relevant to longer software-engineering workflows, which our small bug fix did not measure.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Why did Gemini 3.8 Flash cost more in the tests if the published rate was the same?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Its recorded token usage was higher in every completed pair. A shared per-token rate can still produce a higher total cost when one response consumes more tokens.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Where can I access Gemini 3.8 Flash?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>Google lists the Gemini App, Gemini Enterprise Agent Platform, Google AI Studio, Gemini API, AI Mode, and Google Antigravity. Availability can still vary by account, product, and region.<\/p>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<h3>Are the benchmark results independently verified here?<\/h3>\n<\/div>\n\n\n\n<div class=\"gemini38-review\">\n<p>No. The benchmark section attributes Google&#8217;s official claims directly. Artificial Analysis was unreachable during research, so its leaderboard data is not included as verified evidence in this draft.<\/p>\n<\/div>\n\n\n\n<script type=\"application\/ld+json\">{\n    \"@context\": \"https:\\\/\\\/schema.org\",\n    \"@type\": \"FAQPage\",\n    \"mainEntity\": [\n        {\n            \"@type\": \"Question\",\n            \"name\": \"When was Gemini 3.8 Flash released?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Google DeepMind published the Gemini 3.8 Flash model card on September 2, 2026.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Is Gemini 3.8 Flash generally available?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Yes. Google's official product page lists Gemini 3.8 Flash as generally available.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"What is the Gemini 3.8 Flash API model ID?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"The official model ID is gemini-3.8-flash, which also matched the returned model ID in our API tests.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"How much does Gemini 3.8 Flash cost?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Through December 31, 2026, Google lists $0.75 per 1M input tokens and $3.75 per 1M output tokens, including thinking tokens. Starting January 1, 2027, the rates rise to $1.50 and $7.50.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Is Gemini 3.8 Flash free?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Google's pricing page lists a free tier with limited model access and free input and output tokens. Limits and model availability can differ by account and route.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"What is the Gemini 3.8 Flash context window?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Gemini 3.8 Flash supports up to 1M input tokens and up to 64K tokens of text output, according to the official model card.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Does Gemini 3.8 Flash support images, audio, video, and PDFs?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Yes as inputs. Google's product page lists text, image, video, audio, and PDF inputs; the model card lists text output.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Does Gemini 3.8 Flash support tools?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Google lists function calling, Search as a tool, and computer use. Tool availability can depend on the product and API route.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Is Gemini 3.8 Flash better than Gemini 3.7 Flash?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Not in every task. Gemini 3.8 Flash won one of six usable task pairs in this review, Gemini 3.7 Flash won four, and one pair tied. Google's official evaluations emphasize stronger long-horizon agent and software-engineering performance.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Is Gemini 3.8 Flash good for coding?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Both models produced the correct minimal Python fix in our coding task. Google's official positioning and DeepSWE claim make 3.8 especially relevant to longer software-engineering workflows, which our small bug fix did not measure.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Why did Gemini 3.8 Flash cost more in the tests if the published rate was the same?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Its recorded token usage was higher in every completed pair. A shared per-token rate can still produce a higher total cost when one response consumes more tokens.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Where can I access Gemini 3.8 Flash?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"Google lists the Gemini App, Gemini Enterprise Agent Platform, Google AI Studio, Gemini API, AI Mode, and Google Antigravity. Availability can still vary by account, product, and region.\"\n            }\n        },\n        {\n            \"@type\": \"Question\",\n            \"name\": \"Are the benchmark results independently verified here?\",\n            \"acceptedAnswer\": {\n                \"@type\": \"Answer\",\n                \"text\": \"No. The benchmark section attributes Google's official claims directly. Artificial Analysis was unreachable during research, so its leaderboard data is not included as verified evidence in this draft.\"\n            }\n        }\n    ]\n}<\/script>","protected":false},"excerpt":{"rendered":"<p>Model review \u00b7 Official benchmarks \u00b7 Same-prompt API tests Gemini 3.8 Flash is Google&#8217;s new general-availability Flash model for long-horizon software engineering, agents, and complex knowledge work. The launch claims are strong. Our first controlled tests are much less tidy. This Gemini 3.8 Flash review separates official facts from hands-on evidence. Google&#8217;s pages establish the [&hellip;]<\/p>","protected":false},"author":13,"featured_media":18854,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_seopress_robots_primary_cat":"","_seopress_titles_title":"Gemini 3.8 Flash Review: Pricing, Tests, and Verdict","_seopress_titles_desc":"Is Gemini 3.8 Flash actually better? Compare official claims, API pricing, speed, token use, and real outputs from seven matched tasks.","_seopress_robots_index":"","footnotes":""},"categories":[7],"tags":[],"class_list":["post-18837","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-chat"],"_links":{"self":[{"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/posts\/18837","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/users\/13"}],"replies":[{"embeddable":true,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/comments?post=18837"}],"version-history":[{"count":1,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/posts\/18837\/revisions"}],"predecessor-version":[{"id":18851,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/posts\/18837\/revisions\/18851"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/media\/18854"}],"wp:attachment":[{"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/media?parent=18837"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/categories?post=18837"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/wp.glbgpt.com\/id\/wp-json\/wp\/v2\/tags?post=18837"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}