feat(marketing): surface AgentX across site copy

packages/app/content/blog/ultra-high-interactivity-on-nvidia.mdx

@@ -3,6 +3,7 @@ title: 'Ultra-High Interactivity on NVIDIA GPUs? TileRT on InferenceX'
 subtitle: 'Can TileRT software on NVIDIA GPUs compete with Cerebras, Groq LPU, and SambaNova? Batch size 1, disaggregated engine, high-throughput prefill engine, high-interactivity decode engine'
 date: '2026-08-10'
 publishDate: '2026-08-10'
+modifiedDate: '2026-08-18'
 tags:
   - benchmark
   - gpu
@@ -13,6 +14,8 @@ tags:
   - tilert
   - vllm
   - glm5
+  - agentx
+  - agentic
 ---
 
 _Originally published on the [SemiAnalysis newsletter](https://newsletter.semianalysis.com/p/ultra-high-interactivity-on-nvidia) on August 10, 2026._
@@ -235,9 +238,9 @@ Work is being done to simplify this, especially as software development can be a
 
 AI coding agents accelerate tuning within known templates, but novel transformations still require expert judgment. A monolithic persistent kernel also reduces the usefulness of conventional per-kernel profiler timelines, making automated feedback loops more difficult.
 
-## Next steps with TileRT and InferenceX
+## Next steps with TileRT
 
-We are actively working on moving TileRT benchmarking from InferenceX's single-turn 8k/1k to our new agentic coding benchmark, which we call AgentX. This scenario replays real Claude Code and Codex traces with long-context, multi-turn requests, realistic subagent activity, and dynamic tool-use delays. Its median input length is 140k tokens, while the theoretical median cache-hit rate roofline reaches 99.2%.
+AgentX, InferenceX's long-context, multi-turn coding scenario, is now live for supported models and serving stacks. It replays privacy-preserving workload shapes derived from real Claude Code and Codex sessions, with long-context, multi-turn requests, realistic subagent activity, and dynamic tool-use delays. Its median input length is 140k tokens, while the theoretical median cache-hit rate roofline reaches 99.2%. Extending TileRT benchmarking from its current single-turn 8k/1k workload to AgentX remains the next step for this serving stack.
 
 <Figure
   src="/images/ultra-high-interactivity-on-nvidia/agentx-benchmark-characteristics.png"

packages/app/content/blog/zh/ultra-high-interactivity-on-nvidia.mdx

@@ -3,6 +3,7 @@ title: '在 NVIDIA GPU 上实现超高交互性？TileRT 登陆 InferenceX'
 subtitle: '运行在 NVIDIA GPU 上的 TileRT 软件能否与 Cerebras、Groq LPU、SambaNova 竞争？批大小为 1、分离式引擎、高吞吐量预填充引擎、高交互性解码引擎'
 date: '2026-08-10'
 publishDate: '2026-08-10'
+modifiedDate: '2026-08-18'
 tags:
   - benchmark
   - gpu
@@ -13,6 +14,8 @@ tags:
   - tilert
   - vllm
   - glm5
+  - agentx
+  - agentic
 ---
 
 _本文最初于 2026 年 8 月 10 日发布在 [SemiAnalysis 通讯](https://newsletter.semianalysis.com/p/ultra-high-interactivity-on-nvidia)。_
@@ -235,9 +238,9 @@ TileRT 继承了 ASIC 厂商最大的弱点。静态提前编译意味着模型
 
 AI 编码智能体能在已知模板内加速调优，但新颖的变换仍需专家判断。此外，单体式持久化 kernel 也削弱了传统逐 kernel 性能剖析时间线的作用，使自动化反馈闭环更加困难。
 
-## TileRT 与 InferenceX 的下一步
+## TileRT 的下一步
 
-我们正在积极推进，将 TileRT 的基准测试从 InferenceX 的单轮 8k/1k 场景扩展到我们新的智能体编码基准——AgentX。该场景回放真实的 Claude Code 与 Codex 轨迹，包含长上下文、多轮请求、真实的子智能体活动与动态工具调用延迟。其输入长度中位数为 140k tokens，理论缓存命中率中位数屋顶线可达 99.2%。
+InferenceX 的长上下文多轮编码场景 AgentX 现已在具备对应数据的模型与服务栈上正式上线。它回放从真实 Claude Code 与 Codex 会话衍生出的隐私保护工作负载形态，包含长上下文、多轮请求、真实的 subagent 活动与动态工具调用延迟。其输入长度中位数为 140k tokens，理论缓存命中率中位数屋顶线可达 99.2%。对 TileRT 而言，下一步仍是将目前的单轮 8k/1k 基准测试扩展到 AgentX。
 
 <Figure
   src="/images/ultra-high-interactivity-on-nvidia/agentx-benchmark-characteristics.png"

packages/app/cypress/e2e/overview.cy.ts

@@ -27,8 +27,8 @@ const AGENTX_LABEL_ZH = '长上下文多轮真实智能体场景（AgentX）';
 /** Shared by both locales: the scenario is named after its acronym. */
 const AGENTX_SHORT = 'AgentX';
 
-const PAGE_TITLE = 'Inference Cost per Million Tokens';
-const PAGE_TITLE_ZH = '推理每百万 token 成本';
+const PAGE_TITLE = 'Agentic Inference Costs';
+const PAGE_TITLE_ZH = '智能体推理成本';
 const SOURCE_NOTE = 'Source: InferenceX & SemiAnalysis Market July 2026 AI Cloud TCO Model';
 const SOURCE_LINK_TEXT = 'SemiAnalysis Market July 2026 AI Cloud TCO Model';
 const SOURCE_NOTE_ZH = '来源：InferenceX 与 SemiAnalysis Market July 2026 AI Cloud TCO Model';

packages/app/cypress/e2e/zh-pages.cy.ts

@@ -5,6 +5,7 @@ describe('Chinese (/zh) pages', () => {
     });
 
     it('renders the Chinese landing content', () => {
+      cy.get('[data-testid="intro-section"]').should('contain.text', '智能体推理基准测试');
       cy.contains('h2', '探索 InferenceX').should('exist');
       cy.contains('快速对比').should('exist');
     });
@@ -34,7 +35,7 @@ describe('Chinese (/zh) pages', () => {
     });
 
     it('footer renders in Chinese with zh-internal links', () => {
-      cy.get('[data-testid="footer-brand-description"]').should('contain.text', '开源推理基准测试');
+      cy.get('[data-testid="footer-brand-description"]').should('contain.text', '智能体推理');
       cy.get('[data-testid="footer-link-supporters"]')
         .should('contain.text', '支持者')
         .and('have.attr', 'href', '/zh/quotes');
@@ -62,7 +63,8 @@ describe('Chinese (/zh) pages', () => {
 
     it('renders the Chinese SEO intro above the chart', () => {
       cy.get('[data-testid="zh-tab-intro"]').within(() => {
-        cy.contains('h1', 'AI 推理基准测试').should('exist');
+        cy.contains('h1', '智能体推理基准测试').should('exist');
+        cy.contains('长上下文、多轮').should('exist');
       });
     });
 

packages/app/src/app/about/page.tsx

@@ -21,21 +21,21 @@ const faqJsonLd = {
   })),
 };
 
+const ABOUT_DESCRIPTION =
+  'InferenceX benchmarks agentic and fixed-sequence AI inference across accelerators and serving stacks. AgentX is its long-context, multi-turn coding scenario.';
+
 export const metadata: Metadata = {
   title: 'About',
-  description:
-    'InferenceX is an independent, vendor neutral, reproducible benchmark which continuously benchmarks inference software across a wide range of AI accelerators.',
+  description: ABOUT_DESCRIPTION,
   alternates: enAlternates('/about'),
   openGraph: {
     title: 'About | InferenceX',
-    description:
-      'InferenceX is an independent, vendor neutral, reproducible benchmark which continuously benchmarks inference software across a wide range of AI accelerators.',
+    description: ABOUT_DESCRIPTION,
     url: `${SITE_URL}/about`,
   },
   twitter: {
     title: 'About | InferenceX',
-    description:
-      'InferenceX is an independent, vendor neutral, reproducible benchmark which continuously benchmarks inference software across a wide range of AI accelerators.',
+    description: ABOUT_DESCRIPTION,
   },
 };
 
@@ -47,8 +47,8 @@ export default function AboutPage() {
         <section>
           <Card>
             <h2 className="text-lg font-semibold mb-2">
-              Open Source Continuous Inference Benchmark trusted by Operators of Trillion Dollar
-              GigaWatt Scale Token Factories
+              Open-Source Continuous Agentic Inference Benchmark Trusted by Operators of
+              Trillion-Dollar, GigaWatt-Scale Token Factories
             </h2>
             <p className="text-muted-foreground mb-2">
               As the world progresses exponentially towards AGI, software development and model
@@ -58,9 +58,9 @@ export default function AboutPage() {
             </p>
             <p className="text-muted-foreground mb-2">
               <strong>InferenceX&trade;</strong> (formerly InferenceMAX) is our independent, vendor
-              neutral, reproducible benchmark which addresses these issues by continuously
-              benchmarking inference software across a wide range of AI accelerators that are
-              actually available to the ML community.
+              neutral, reproducible benchmark. It measures both fixed-sequence serving and AgentX,
+              our long-context, multi-turn agentic coding workload, across AI accelerators and
+              serving stacks available to the ML community.
             </p>
             <p className="text-muted-foreground">
               Our open data & insights are widely adopted by the ML community, capacity planning

packages/app/src/app/blog/page.tsx

@@ -11,11 +11,11 @@ import { SITE_URL, SITE_NAME, AUTHOR_NAME } from '@semianalysisai/inferencex-con
 
 export const metadata: Metadata = {
   title: 'Articles',
-  description: `Technical articles from ${SITE_NAME} by ${AUTHOR_NAME} — AI inference benchmarking, chip performance analysis, and ML infrastructure insights.`,
+  description: `Technical articles from ${SITE_NAME} by ${AUTHOR_NAME} on agentic inference benchmarks, AgentX results, chip performance, and ML infrastructure.`,
   alternates: enAlternates('/blog'),
   openGraph: {
     title: `Articles | ${SITE_NAME} by ${AUTHOR_NAME}`,
-    description: 'AI inference benchmarking insights and chip performance analysis.',
+    description: 'Articles on agentic inference benchmarks, AgentX results, and chip performance.',
     url: `${SITE_URL}/blog`,
   },
 };
@@ -49,7 +49,8 @@ export default async function BlogPage({
           <Card>
             <h2 className="text-2xl lg:text-4xl font-bold tracking-tight">Articles</h2>
             <p className="mt-3 text-base lg:text-lg text-muted-foreground">
-              Insights on AI inference benchmarking, chip performance, and ML infrastructure.
+              Articles on agentic inference, AgentX results, chip performance, and ML
+              infrastructure.
             </p>
             <p className="text-sm text-muted-foreground">
               New to the terminology?{' '}

packages/app/src/app/compare/page.tsx

@@ -19,7 +19,7 @@ import { bucketComparePairsByVendor, formatModelList } from '@/lib/compare-ssr';
 
 export const dynamic = 'force-dynamic';
 
-const DESCRIPTION = `InferenceX is the independent, open-source chip inference benchmark from SemiAnalysis, with verified, reproducible nightly results. ${SUPPORTERS_LINE} Compare latency, throughput & cost head-to-head across DeepSeek V4 Pro, DeepSeek R1, Kimi K2, MiniMax M3, GLM 5, Qwen 3.5 & more.`;
+const DESCRIPTION = `InferenceX is the independent, open-source chip inference benchmark from SemiAnalysis, with verified, reproducible results updated as configurations change. ${SUPPORTERS_LINE} Compare latency, throughput & cost head-to-head across DeepSeek V4 Pro, DeepSeek R1, Kimi K2, MiniMax M3, GLM 5, Qwen 3.5 & more.`;
 
 export const metadata: Metadata = {
   title: 'Chip Comparisons',

packages/app/src/app/layout.tsx

@@ -54,6 +54,11 @@ export const metadata: Metadata = {
   },
   description: DESCRIPTION,
   keywords: [
+    'AgentX scenario',
+    'agentic inference benchmark',
+    'agentic coding benchmark',
+    'long context inference benchmark',
+    'multi-turn inference benchmark',
     'AI inference benchmark',
     'GPU benchmark',
     'LLM benchmark',
@@ -160,7 +165,7 @@ const jsonLd = {
       '@id': `${SITE_URL}/#application`,
       name: SITE_NAME,
       description:
-        'Open-source AI inference benchmark dashboard. Compare chip performance for LLM inference across NVIDIA GB200, H100, AMD MI355X, and more.',
+        'InferenceX is an open-source agentic inference benchmark dashboard. It compares the AgentX long-context, multi-turn coding scenario with fixed-sequence serving across NVIDIA, AMD, and other accelerators.',
       url: SITE_URL,
       applicationCategory: 'DeveloperApplication',
       operatingSystem: 'Web',

packages/app/src/app/llms-full.txt/route.ts

@@ -28,7 +28,7 @@ export async function GET() {
     `# ${SITE_NAME} Articles — Full Content`,
     `> By ${AUTHOR_NAME}`,
     '',
-    `This file contains the full text of all articles from ${SITE_NAME} (${SITE_URL}/blog).`,
+    `This file contains the full text of all articles from ${SITE_NAME} (${SITE_URL}/blog), covering agentic inference benchmarks including AgentX results, fixed-sequence AI inference, chip performance, and ML infrastructure.`,
     `It is intended for consumption by large language models and AI assistants.`,
     '',
     '---',

packages/app/src/app/llms.txt/route.ts

@@ -8,7 +8,7 @@ export async function GET() {
   const lines = [
     `# ${SITE_NAME} by ${AUTHOR_NAME}`,
     '',
-    `> ${SITE_NAME} is the open-source AI inference benchmark dashboard. We compare chip performance for LLM inference across NVIDIA GB200, H100, AMD MI355X, and more.`,
+    `> ${SITE_NAME} is an open-source agentic inference benchmark dashboard. It compares the AgentX long-context, multi-turn coding scenario with fixed-sequence serving across NVIDIA, AMD, and other accelerators using public runs.`,
     '',
     `## Links`,
     '',

packages/app/src/app/manifest.ts

@@ -7,7 +7,7 @@ export default function manifest(): MetadataRoute.Manifest {
     name: `${SITE_NAME} by ${AUTHOR_NAME}`,
     short_name: SITE_NAME,
     description:
-      'Open-source AI inference benchmark. Compare chip performance across NVIDIA, AMD, and more.',
+      'Open-source agentic inference benchmark. Compare the AgentX scenario and fixed-sequence serving performance across NVIDIA, AMD, and more.',
     start_url: '/',
     display: 'standalone',
     background_color: '#09090b',

packages/app/src/app/overview/page.tsx

@@ -18,21 +18,21 @@ import { getOverviewPageData } from '@/lib/overview-data.server';
 export const dynamic = 'force-dynamic';
 
 const DESCRIPTION =
-  'Compare hyperscaler cost per million total tokens across MI355X, B200, B300, GB200 and GB300 using the scenario shown for each active model.';
+  'Compare hyperscaler cost per million total tokens across MI355X, B200, B300, GB200, and GB300 for the AgentX long-context, multi-turn coding scenario and fixed-sequence scenarios where data is available.';
 
 export const metadata: Metadata = {
-  title: 'Inference Cost Overview',
+  title: 'Agentic Inference Costs',
   description: DESCRIPTION,
   alternates: enAlternates('/overview'),
   openGraph: {
-    title: `Inference Cost Overview | ${SITE_NAME}`,
+    title: `Agentic Inference Costs | ${SITE_NAME}`,
     description: DESCRIPTION,
     url: `${SITE_URL}/overview`,
     type: 'website',
   },
   twitter: {
     card: 'summary_large_image',
-    title: `Inference Cost Overview | ${SITE_NAME}`,
+    title: `Agentic Inference Costs | ${SITE_NAME}`,
     description: DESCRIPTION,
   },
 };

packages/app/src/app/zh/about/page.tsx

@@ -22,22 +22,22 @@ const faqJsonLd = {
   })),
 };
 
+const ABOUT_DESCRIPTION =
+  'InferenceX 对比各类加速器与服务栈的智能体推理和固定序列 AI 推理性能。AgentX 是其长上下文多轮编码场景。';
+
 export const metadata: Metadata = {
   title: '关于',
-  description:
-    'InferenceX 是一个独立、厂商中立、可复现的基准测试平台，持续测试各类 AI 加速器上的推理软件性能。',
+  description: ABOUT_DESCRIPTION,
   alternates: zhAlternates('/about'),
   openGraph: {
     title: '关于 | InferenceX',
-    description:
-      'InferenceX 是一个独立、厂商中立、可复现的基准测试平台，持续测试各类 AI 加速器上的推理软件性能。',
+    description: ABOUT_DESCRIPTION,
     url: `${SITE_URL}/zh/about`,
     locale: ZH_OG_LOCALE,
   },
   twitter: {
     title: '关于 | InferenceX',
-    description:
-      'InferenceX 是一个独立、厂商中立、可复现的基准测试平台，持续测试各类 AI 加速器上的推理软件性能。',
+    description: ABOUT_DESCRIPTION,
   },
 };
 
@@ -49,16 +49,16 @@ export default function AboutPageZh() {
         <section>
           <Card>
             <h2 className="text-lg font-semibold mb-2">
-              开源持续推理基准测试——受万亿美元级吉瓦规模 Token 工厂运营者的信赖
+              开源持续智能体推理基准测试——受万亿美元级吉瓦规模 Token 工厂运营者的信赖
             </h2>
             <p className="text-muted-foreground mb-2">
               随着世界以指数级速度迈向
               AGI，软件开发和模型发布日新月异。现有基准测试因其静态性质而迅速过时，参与者往往提交专为基准测试定制的软件镜像，无法反映真实的线上推理性能。
             </p>
             <p className="text-muted-foreground mb-2">
               <strong>InferenceX&trade;</strong>（原名
-              InferenceMAX）是我们独立、厂商中立、可复现的基准测试平台，通过持续测试实际可用于 ML
-              社区的各类 AI 加速器上的推理软件来解决这些问题。
+              InferenceMAX）是我们独立、厂商中立、可复现的基准测试平台。它测试固定序列推理服务和
+              AgentX 长上下文多轮智能体编码工作负载，覆盖 ML 社区实际可用的各类 AI 加速器与服务栈。
             </p>
             <p className="text-muted-foreground">
               我们的开放数据与洞察已被 ML 社区广泛采用，包括万亿美元级 Token 工厂和 AI

packages/app/src/app/zh/blog/page.tsx

@@ -11,11 +11,11 @@ import { SITE_URL, SITE_NAME, AUTHOR_NAME } from '@semianalysisai/inferencex-con
 
 export const metadata: Metadata = {
   title: '文章',
-  description: `${SITE_NAME} by ${AUTHOR_NAME} 的技术文章——AI 推理基准测试、Chip 性能分析与 ML 基础设施洞见。`,
+  description: `${SITE_NAME} by ${AUTHOR_NAME} 发布关于智能体推理基准测试、AgentX 结果、Chip 性能与 ML 基础设施的技术文章。`,
   alternates: zhAlternates('/blog'),
   openGraph: {
     title: `文章 | ${SITE_NAME} by ${AUTHOR_NAME}`,
-    description: 'AI 推理基准测试洞见与 Chip 性能分析。',
+    description: '关于智能体推理基准测试、AgentX 结果与 Chip 性能的文章。',
     url: `${SITE_URL}/zh/blog`,
     locale: ZH_OG_LOCALE,
   },
@@ -51,7 +51,7 @@ export default async function ZhBlogPage({
           <Card>
             <h2 className="text-2xl lg:text-4xl font-bold tracking-tight">文章</h2>
             <p className="mt-3 text-base lg:text-lg text-muted-foreground">
-              关于 AI 推理基准测试、Chip 性能与 ML 基础设施的深度洞见。
+              关于智能体推理、AgentX 结果、Chip 性能与 ML 基础设施的文章。
             </p>
             <p className="text-sm text-muted-foreground">
               不熟悉相关概念？{' '}

packages/app/src/app/zh/compare/page.tsx

@@ -18,7 +18,7 @@ import { ZH_OG_LOCALE, zhAlternates } from '@/lib/i18n';
 
 export const dynamic = 'force-dynamic';
 
-const DESCRIPTION = `InferenceX 是 SemiAnalysis 推出的独立开源 Chip 推理基准测试平台，提供经过验证的、可复现的每夜测试结果。${SUPPORTERS_LINE_ZH}横向对比 DeepSeek V4 Pro、DeepSeek R1、Kimi K2、MiniMax M3、GLM 5、Qwen 3.5 等模型的延迟、吞吐量与成本。`;
+const DESCRIPTION = `InferenceX 是 SemiAnalysis 推出的独立开源 Chip 推理基准测试平台，提供经过验证、可复现并随配置变化更新的测试结果。${SUPPORTERS_LINE_ZH}横向对比 DeepSeek V4 Pro、DeepSeek R1、Kimi K2、MiniMax M3、GLM 5、Qwen 3.5 等模型的延迟、吞吐量与成本。`;
 
 export const metadata: Metadata = {
   title: 'Chip 对比',