{"id":13268,"date":"2026-06-02T15:13:30","date_gmt":"2026-06-02T07:13:30","guid":{"rendered":"https:\/\/ai-stack.ai\/?p=13268"},"modified":"2026-06-03T16:37:20","modified_gmt":"2026-06-03T08:37:20","slug":"claude-opus-4-8","status":"publish","type":"post","link":"https:\/\/ai-stack.ai\/en\/claude-opus-4-8","title":{"rendered":"Claude Opus 4.8 Explained: Anthropic\u2019s First Model That Says \u201cI\u2019m Not Sure\u201d \u2014 Agentic Coding Redefines the Enterprise AI Baseline"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\">Introduction: Six Weeks, One Leap<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">On May 28, 2026, Anthropic released Claude Opus 4.8 \u2014 just six weeks after Opus 4.7 launched on April 16. With GPT-5.5 arriving on April 23 and Gemini 3.1 Pro Preview surfacing in May, the iteration cadence in frontier AI has never been this compressed.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">But the headline benchmarks only tell part of the story. Opus 4.8 marks three qualitative shifts that matter more to enterprises than any single benchmark score:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>First, it\u2019s the first frontier model that can genuinely say \u201cI\u2019m not sure\u201d instead of fabricating a plausible-sounding answer.<\/strong> Anthropic reports Opus 4.8 is roughly four times less likely than Opus 4.7 to let code flaws pass unremarked.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Second, it achieves a 69.2% score on SWE-bench Pro \u2014 a 10.6-point gap over GPT-5.5\u2019s 58.6%.<\/strong> This is the widest lead in agentic coding that any publicly available model has held.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Third, Dynamic Workflows enable a single Claude session to spin up hundreds of parallel sub-agents,<\/strong> coordinating large-scale tasks like codebase migrations across hundreds of thousands of lines \u2014 from kickoff to merge.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This article analyzes Opus 4.8 through the lens of enterprise AI infrastructure: what the benchmarks mean, how the pricing works, and what the shift toward agentic workflows demands of your compute strategy.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">I. The Numbers: What Changed in Six Weeks<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">1.1 Agentic Coding: A 10.6-Point Gap Opens Up<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Benchmark<\/th><th>Claude Opus 4.8<\/th><th>GPT-5.5<\/th><th>Claude Mythos (preview)<\/th><\/tr><\/thead><tbody><tr><td>SWE-bench Pro<\/td><td>69.2%<\/td><td>58.6%<\/td><td>77.8%<\/td><\/tr><tr><td>SWE-bench Verified<\/td><td>88.6%<\/td><td>\u2014<\/td><td>\u2014<\/td><\/tr><tr><td>Terminal-Bench 2.1<\/td><td>74.6%<\/td><td>78.2%<\/td><td>\u2014<\/td><\/tr><tr><td>HLE (no tools)<\/td><td>49.8%<\/td><td>41.4%<\/td><td>64.7%<\/td><\/tr><tr><td>HLE (with tools)<\/td><td>57.9%<\/td><td>52.2%<\/td><td>\u2014<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Sources: Anthropic official release, Artificial Analysis independent testing, R&amp;D World comparison<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The SWE-bench Pro gap is the headline figure. But Terminal-Bench 2.1 tells a more nuanced story: GPT-5.5 leads at 78.2% vs.&nbsp;74.6%, and Oracle\u2019s own tests show GPT-5.5 reaching 83.4% under the Codex CLI harness. The takeaway: if your engineering workload is shell-heavy infrastructure automation, GPT-5.5 retains an edge. If it\u2019s codebase-scale software engineering \u2014 multi-file refactors, large-scale migrations, collaborative editing \u2014 Opus 4.8\u2019s lead is unambiguous.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1.2 Knowledge Work: GDPval-AA Elo Hits 1,890<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Opus 4.8 scores 1,890 on GDPval-AA Elo vs.&nbsp;GPT-5.5\u2019s 1,769 \u2014 a 121-point gap that translates to roughly a 67% head-to-head win rate (source: Anthropic official GDPval-AA dataset). On Humanity\u2019s Last Exam, Opus 4.8 leads in both tool-free (49.8% vs.&nbsp;41.4%) and tool-augmented (57.9% vs.&nbsp;52.2%) configurations.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1.3 Computer Use<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">On OSWorld-Verified, Opus 4.8 scores 83.4% vs.&nbsp;GPT-5.5\u2019s 78.7%. On Online-Mind2Web, it hits 84%, which Anthropic describes as \u201ca meaningful jump over both Opus 4.7 and GPT-5.5\u201d (source: Anthropic official release).<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1.4 Third-Party Chinese-Language Benchmarks<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">SuperCLUE\u2019s May 30 evaluation placed Opus 4.8 at #1 globally in three categories (source: SuperCLUE Chinese benchmark):<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Domain<\/th><th>Score<\/th><th>Global Rank<\/th><\/tr><\/thead><tbody><tr><td>Code Generation<\/td><td>83.58<\/td><td>#1<\/td><\/tr><tr><td>Hallucination Control<\/td><td>87.48<\/td><td>#1<\/td><\/tr><tr><td>Scientific Reasoning<\/td><td>77.19<\/td><td>#1<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">The composite score of 73.93 places Opus 4.8 in the same tier as GPT-5.5 and Gemini 3.1 Pro Preview. However, SuperCLUE noted a \u201crelatively obvious\u201d decline in complex instruction-following \u2014 which means enterprises should test Opus 4.8 against their specific multi-step workflows before deploying. For example: generating brand-compliant business presentations in a specific format (competitor analysis, brand defense strategy reports), or producing legal documents that must strictly adhere to the same compliance framework across multiple rounds of revision \u2014 these are the kinds of scenarios where instruction-following regressions could surface.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For a deeper look at how the Opus line has evolved, see our <a href=\"https:\/\/ai-stack.ai\/en\/claude-opus-4-5\">Claude Opus 4.5 enterprise deployment guide<\/a>; for a head-to-head selection framework, refer to <a href=\"https:\/\/ai-stack.ai\/en\/claude-opus-4-6-vs-gpt-5-3-codex2026\">Claude Opus 4.6 vs.&nbsp;GPT-5.3: 2026 AI Model Selection Guide<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">II. Dynamic Workflows: One Claude, Hundreds of Sub-Agents<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">2.1 How It Works<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Dynamic Workflows, available as a research preview in Claude Code, lets Opus 4.8 plan a task and then spawn parallel sub-agents to execute it. Key specs (source: Anthropic official):<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Up to 1,000 sub-agents<\/strong> per session<\/li>\n\n\n\n<li><strong>16 concurrent<\/strong> sub-agents at any time<\/li>\n\n\n\n<li><strong>Extended runtimes<\/strong>: sub-agents can work on longer tasks without timing out<\/li>\n\n\n\n<li><strong>Self-verification<\/strong>: sub-agents check their outputs before reporting back<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">2.2 The Enterprise Shift: From \u201cWrite This Function\u201d to \u201cMigrate This Codebase\u201d<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Early tester reports describe Opus 4.8 handling codebase-scale migrations \u2014 language rewrites, monorepo dependency refactors, batch test generation across hundreds of files \u2014 in a single session. This is fundamentally different from the \u201ccopilot\u201d paradigm. The model isn\u2019t assisting one developer; it\u2019s functioning as a distributed engineering team.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Dynamic Workflows is currently available on Claude Code Enterprise, Team, and Max plans.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">From sub-agent coordination to multi-step autonomous planning, the engineering of AI agents is evolving rapidly. \ud83d\udd17 Further Reading: <a href=\"https:\/\/ai-stack.ai\/en\/ai-agent-development\">The Reality of AI Agent Development: From Single API to Complex Systems<\/a>, where we trace the technical path from monolithic models to multi-agent architectures and enterprise adoption considerations.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">2.3 The Hidden Infrastructure Implication<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Dynamic Workflows radically changes token consumption patterns. A single task that spawns 200 sub-agents, each consuming tens of thousands of tokens, can burn through millions of tokens \u2014 orders of magnitude more than a standard chat interaction. This means:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Per-seat budgeting breaks.<\/strong> Cost models must shift to task-level tracking.<\/li>\n\n\n\n<li><strong>Rate limits become a bottleneck.<\/strong> When multiple teams trigger large workflows simultaneously, API rate limits will gate throughput.<\/li>\n\n\n\n<li><strong>GPU scheduling matters more than GPU count.<\/strong> For enterprises running on-premise models alongside cloud APIs, the ability to dynamically allocate GPU resources across teams and tasks becomes the ROI bottleneck \u2014 not the total number of GPUs.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">III. Effort Control: Thinking Depth as a Cost Variable<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Opus 4.8 introduces five effort levels on claude.ai and Cowork (source: Anthropic official):<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Level<\/th><th>Label<\/th><th>Best For<\/th><\/tr><\/thead><tbody><tr><td>Low<\/td><td>low<\/td><td>Quick lookups, format conversion<\/td><\/tr><tr><td>Auto<\/td><td>auto<\/td><td>General conversation<\/td><\/tr><tr><td>High (default)<\/td><td>high<\/td><td>Daily coding, writing, analysis<\/td><\/tr><tr><td>Extra<\/td><td>xhigh<\/td><td>Complex refactors, async workflows<\/td><\/tr><tr><td>Max<\/td><td>max<\/td><td>Mission-critical reasoning<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Default is High, with token cost comparable to Opus 4.7\u2019s default \u2014 meaning you get better performance at the same price.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The enterprise playbook: route simple queries through Low, use High for daily engineering work, reserve Extra\/Max for tasks where errors are costly. This makes \u201cthinking depth\u201d a tunable cost parameter rather than a black-box decision.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">IV. The Honesty Revolution: \u201cI\u2019m Not Sure\u201d as a Feature<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">4.1 Four Times Less Likely to Let Flaws Slip Through<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Anthropic\u2019s most underrated claim: Opus 4.8 is \u201caround four times less likely than its predecessor to allow flaws in code it has written to pass unremarked\u201d (source: Anthropic official release). Early testers confirm the model \u201cproactively flags issues with the inputs and outputs of an analysis, something other models routinely miss\u201d (source: Anthropic-cited tester Michael Ran).<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">4.2 Why This Matters for Enterprises<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">A model that confidently delivers wrong code costs far more than one that says \u201cI\u2019m not sure about this.\u201d In regulated industries \u2014 finance, healthcare, legal \u2014 an uncaught AI error in production can trigger compliance violations, financial loss, or worse. Opus 4.8\u2019s honesty improvement means enterprises can begin to build trust mechanisms around what the model <em>refuses to claim<\/em> rather than just what it generates.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">4.3 Alignment Progress<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Anthropic\u2019s Alignment team reports Opus 4.8 \u201creaches new highs on measures of prosocial traits\u201d with misalignment rates \u201csubstantially lower than Opus 4.7\u201d and alignment quality \u201csimilar to our best-aligned model, Claude Mythos Preview\u201d (source: Anthropic Opus 4.8 System Card). For regulated enterprises, this is becoming a procurement factor \u2014 not just \u201chow smart is the model\u201d but \u201chow safe is it.\u201d<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">V. Fast Mode: 2.5\u00d7 Faster, 3\u00d7 Cheaper<\/h2>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Mode<\/th><th>Input (per 1M tokens)<\/th><th>Output (per 1M tokens)<\/th><\/tr><\/thead><tbody><tr><td>Standard<\/td><td>$5.00<\/td><td>$25.00<\/td><\/tr><tr><td>Fast Mode<\/td><td>$10.00<\/td><td>$50.00<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">API model ID: claude-opus-4-8. Available on Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Fast Mode\u2019s 3\u00d7 price reduction makes low-latency inference financially viable for production workloads \u2014 real-time customer support, interactive analytics, live coding assistance. The trade-off to calculate: while Fast Mode is now 3\u00d7 cheaper than its predecessor, its output cost is still 2\u00d7 the current standard mode ($50 vs $25). In other words, you\u2019re paying 2\u00d7 the output price for 2.5\u00d7 the speed. If latency doesn\u2019t matter (batch reporting, offline data processing), standard mode is more economical; if latency is critical (live customer support, real-time coding assistance), Fast Mode is now affordable enough to be the default. The key is not defaulting every task to the same mode \u2014 treat mode selection as a cost-control lever.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">VI. Beyond Opus 4.8: The Mythos Horizon<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Anthropic confirmed that Mythos-class models will be available to all customers \u201cin the coming weeks\u201d (source: Anthropic official). Mythos Preview currently scores 77.8% on SWE-bench Pro and 64.7% on HLE with tools, and is restricted to Project Glasswing cybersecurity partners. The dual-track strategy is clear: Opus iterates fast and ships to everyone; Mythos pushes the frontier under tighter safety controls before broader release.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For enterprise buyers, the message is: the capability curve is still steep. Don\u2019t optimize procurement for \u201cwho\u2019s winning today\u201d \u2014 optimize for iteration velocity, safety track record, and ecosystem stability.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">VII. What This Means for Enterprise AI Infrastructure<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">7.1 Stop Asking \u201cWhich Model Wins.\u201d Start Asking \u201cWhich Model Does What.\u201d<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Opus 4.8 leads agentic coding. GPT-5.5 leads terminal-heavy automation. Gemini has different strengths. No single model dominates every benchmark.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The operational answer is multi-model routing: for example, Opus 4.8 dominates SWE-bench (ideal for large-scale refactors and multi-file collaborative editing), but trails GPT-5.5 on Terminal-Bench \u2014 and that gap directly tells you the division of labor: Opus 4.8 for software engineering, GPT-5.5 for shell automation and infrastructure scripting, open-source models for sensitive on-premise data \u2014 all managed through a unified infrastructure layer that handles routing, quotas, cost tracking, and access control. No single model wins everywhere, but the combination is a clean sweep.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">7.2 GPU Utilization Is the Real ROI Variable<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Model generations ship every six weeks. GPU hardware cycles are 3\u20135 years. These timelines don\u2019t match. The variable that determines ROI isn\u2019t \u201chow many GPUs do we own\u201d \u2014 it\u2019s \u201cwhat percentage of our GPU hours are actually utilized across teams, tasks, and models.\u201d Platforms that provide GPU partitioning (MIG\/vGPU), multi-tenant management, and dynamic scheduling become the difference between a cost center and a productivity multiplier.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\ud83d\udd17 Further Reading: <a href=\"https:\/\/ai-stack.ai\/en\/manage-gpu-effectively\">How to Manage GPU Resources Effectively for Enterprise AI<\/a> dives into the technical details of GPU partitioning and multi-tenant orchestration; <a href=\"https:\/\/ai-stack.ai\/en\/gtc-2026-nemoclaw\">GTC 2026 Complete Analysis: NemoClaw as the New Enterprise Agent OS Standard<\/a> explores the infrastructure implications of agentic AI from the Agent OS perspective.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">7.3 Token Costs Need Task-Level Granularity<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">When a single Dynamic Workflow can consume millions of tokens, aggregate monthly API bills are useless. You need to track which team, which use case, and which effort level is driving consumption.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">7.4 Honesty Changes the Trust Equation<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">When a model can say \u201cI\u2019m not sure,\u201d enterprises need workflows that handle those moments \u2014 who verifies, what triggers human review, and how the decision gets logged. This is a governance question, not an engineering one.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Conclusion: The Roadmark, Not the Destination<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Opus 4.8 isn\u2019t just a faster model. It\u2019s a signal that AI is transitioning across four structural shifts:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>From answering questions to executing tasks<\/strong> \u2014 Dynamic Workflows turn AI from a passive responder into an active coordinator<\/li>\n\n\n\n<li><strong>From always-confident to appropriately-uncertain<\/strong> \u2014 honesty becomes a measurable model quality<\/li>\n\n\n\n<li><strong>From single-model bets to multi-model routing<\/strong> \u2014 enterprise competitiveness lives in the orchestration layer<\/li>\n\n\n\n<li><strong>From \u201chow smart\u201d to \u201chow safe\u201d<\/strong> \u2014 alignment quality enters procurement criteria<\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">The practical takeaway for enterprises evaluating or deploying AI: the models will keep getting better every six weeks. What won\u2019t change is the need for a compute governance layer \u2014 GPU scheduling, cost tracking at task granularity, multi-model routing, and security compliance \u2014 that can absorb whatever model comes next.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Opus 4.8, Mythos, GPT-6 \u2014 whichever one wins, they all need the same enterprise infrastructure underneath.<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Anthropic launched Claude Opus 4.8 on May 28, 2026, achieving 69.2% on SWE-bench Pro and introducing Dynamic Workflows with up to 1,000 parallel sub-agents. This analysis covers the benchmark landscape, the honesty revolution (4\u00d7 fewer uncaught code flaws), Effort Control pricing strategy, and what these shifts mean for enterprise GPU infrastructure and multi-model deployment.<\/p>\n","protected":false},"author":253372376,"featured_media":13294,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_crdt_document":"","jetpack_post_was_ever_published":true,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[96987604,96987592],"tags":[96988710,96988709,96988708],"class_list":["post-13268","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-news","category-featured-articles","tag-claude-opus-4-8","tag-anthropic","tag-opus-4-8"],"blocksy_meta":[],"acf":[],"jetpack_featured_media_url":"https:\/\/i0.wp.com\/ai-stack.ai\/wp-content\/uploads\/2026\/06\/en-e47d4f88.jpg?fit=1920%2C1080&quality=100&ct=202603031250&ssl=1","jetpack_shortlink":"https:\/\/wp.me\/ph344V-3s0","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/posts\/13268","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/users\/253372376"}],"replies":[{"embeddable":true,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/comments?post=13268"}],"version-history":[{"count":2,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/posts\/13268\/revisions"}],"predecessor-version":[{"id":13276,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/posts\/13268\/revisions\/13276"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/media\/13294"}],"wp:attachment":[{"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/media?parent=13268"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/categories?post=13268"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ai-stack.ai\/en\/wp-json\/wp\/v2\/tags?post=13268"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}