[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EMG7QFXGRPBH365J4ECPPD":3,"skill-same:01M7EMG7QFXGRPBH365J4ECPPD":47},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":6,"updatedAt":32,"install":33,"resources":46},"01M7EMG7QFXGRPBH365J4ECPPD","cerebras-guide",null,"Blazing inference on Cerebras wafer-scale hardware — ultra-fast open-model APIs for speed-critical workloads.","TanNhatCMS\u002Fjarvis","TanNhatCMS","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F113796420?v=4","01M753MV17KG5CD3152G3WQEQY","KietNT","01M7EKEDYVYQ6PWVJJMZ911VMX",0,"MIT","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fams-cerebras-guide\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: cerebras-guide\ndescription: Blazing inference on Cerebras wafer-scale hardware — ultra-fast open-model APIs for speed-critical workloads.\ncategory: ai-research\n---\n\n## Overview\n\nCerebras builds wafer-scale AI chips — entire processors the size of a silicon\nwafer — and offers inference APIs running open models on that hardware at\nextreme speed. Like Groq, the proposition is hardware-level velocity: very high\ntokens-per-second and low latency for supported models, delivered via a simple\nAPI.\n\nCerebras competes in the same \"speed specialist\" category as Groq, with its own\nhardware architecture and model set. The practical approach is to evaluate both\n(and other speed-optimized providers) on your workload: supported models,\nmeasured latency on your prompts, pricing, and rate limits. The winner is\nworkload-specific, not universal.\n\nUse Cerebras when latency or throughput is the binding constraint and a\nCerebras-supported model meets your quality bar. It's a specialist instrument —\nevaluate it as one.\n\n## When to use\n\n- Speed-critical inference: real-time agents, voice, interactive copilots.\n- High-throughput generation where tokens\u002Fsec per dollar is the metric.\n- Agent loops with sequential calls that compound latency.\n- Benchmarking speed-optimized providers against each other.\n- Workloads where a Cerebras-supported model clears your quality bar.\n- Throughput-heavy batch jobs on supported models.\n\n## Core concepts\n\n- **Wafer-scale hardware**: Cerebras chips are entire wafers — massive on-chip\n  memory and bandwidth that eliminate the memory bottlenecks limiting GPU\n  inference speed. The architecture is the advantage.\n- **Supported model set**: curated models compiled for Cerebras hardware. Check\n  availability for your needs — curated, not comprehensive.\n- **Extreme throughput**: headline tokens\u002Fsec figures among the highest\n  available. Validate on your prompt shapes and context lengths.\n- **Low-latency serving**: fast time-to-first-token plus fast generation —\n  both matter for interactive use; measure both.\n- **API access**: straightforward API for supported models; evaluate integration\n  effort (usually minimal for OpenAI-compatible surfaces).\n- **Pricing model**: per-token pricing for the speed tier. Model cost per\n  completed task against alternatives — speed has a price; verify it's worth it.\n- **Rate limits**: understand concurrency and throughput limits for your\n  production shape. Speed without sufficient quota doesn't ship.\n- **Model updates**: supported models and versions evolve; track changes and\n  re-validate quality on updates.\n\n## Practical workflow\n\n1. **Check model support first.** Confirm a quality-sufficient model is\n   available. No supported model meeting your bar → no engagement, regardless\n   of speed.\n2. **Benchmark head-to-head.** Same eval set, same prompts: Cerebras vs. your\n   current provider (and vs. other speed specialists). Measure quality, TTFT,\n   tokens\u002Fsec, and cost per task.\n3. **Test your context lengths.** Wafer-scale advantages can vary with sequence\n   length. Benchmark at your real context sizes, including your p99.\n4. **Validate output behavior.** Run your full eval suite — confirm the\n   hardware-optimized serving preserves the quality and behavior you need.\n5. **Model the economics.** Per-token price × tokens per task ÷ success rate =\n   cost per completed task. Compare honestly against slower, cheaper options.\n6. **Plan capacity and fallbacks.** Map rate limits to production traffic;\n   configure fallback providers for resilience.\n7. **Monitor production performance.** Track latency percentiles and quality\n   signals on live traffic; alert on drift.\n\nChecklist for Cerebras in production:\n- Supported model validated on your eval set.\n- Head-to-head benchmark vs. alternatives (quality + speed + cost).\n- Context-length behavior tested at your p99.\n- Rate limits mapped to production needs; fallbacks configured.\n- Cost per completed task modeled and acceptable.\n\n## Common pitfalls\n\n- **Speed as the only criterion.** Ultra-fast inference of inadequate quality\n  is worthless. Quality bar first, always.\n- **Headline-number anchoring.** Marketing tokens\u002Fsec measured under ideal\n  conditions. Your prompts, your context lengths, your concurrency — measure\n  those.\n- **Model availability assumptions.** Building plans around a model before\n  confirming it's supported. Check first.\n- **No head-to-head evaluation.** Choosing on reputation rather than measured\n  comparison. The speed-specialist field is competitive — benchmark.\n- **Ignoring the cost per task.** Faster per-token pricing can still lose on\n  cost per completed task if it uses more tokens or fails more often.\n- **Single-provider fragility.** Production on one speed specialist without\n  fallbacks. Fast outages are still outages.\n- **Context-length blind spots.** Validating speed at 1k tokens, deploying at\n  32k. Performance characteristics change with length — test yours.\n- **Skipping re-validation on updates.** Supported model versions change;\n  behavior can shift. Re-run evals when versions move.\n","## Overview\n\nCerebras builds wafer-scale AI chips — entire processors the size of a silicon\nwafer — and offers inference APIs running open models on that hardware at\nextreme speed. Like Groq, the proposition is hardware-level velocity: very high\ntokens-per-second and low latency for supported models, delivered via a simple\nAPI.\n\nCerebras competes in the same \"speed specialist\" category as Groq, with its own\nhardware architecture and model set. The practical approach is to evaluate both\n(and other speed-optimized providers) on your workload: supported models,\nmeasured latency on your prompts, pricing, and rate limits. The winner is\nworkload-specific, not universal.\n\nUse Cerebras when latency or throughput is the binding constraint and a\nCerebras-supported model meets your quality bar. It's a specialist instrument —\nevaluate it as one.\n\n## When to use\n\n- Speed-critical inference: real-time agents, voice, interactive copilots.\n- High-throughput generation where tokens\u002Fsec per dollar is the metric.\n- Agent loops with sequential calls that compound latency.\n- Benchmarking speed-optimized providers against each other.\n- Workloads where a Cerebras-supported model clears your quality bar.\n- Throughput-heavy batch jobs on supported models.\n\n## Core concepts\n\n- **Wafer-scale hardware**: Cerebras chips are entire wafers — massive on-chip\n  memory and bandwidth that eliminate the memory bottlenecks limiting GPU\n  inference speed. The architecture is the advantage.\n- **Supported model set**: curated models compiled for Cerebras hardware. Check\n  availability for your needs — curated, not comprehensive.\n- **Extreme throughput**: headline tokens\u002Fsec figures among the highest\n  available. Validate on your prompt shapes and context lengths.\n- **Low-latency serving**: fast time-to-first-token plus fast generation —\n  both matter for interactive use; measure both.\n- **API access**: straightforward API for supported models; evaluate integration\n  effort (usually minimal for OpenAI-compatible surfaces).\n- **Pricing model**: per-token pricing for the speed tier. Model cost per\n  completed task against alternatives — speed has a price; verify it's worth it.\n- **Rate limits**: understand concurrency and throughput limits for your\n  production shape. Speed without sufficient quota doesn't ship.\n- **Model updates**: supported models and versions evolve; track changes and\n  re-validate quality on updates.\n\n## Practical workflow\n\n1. **Check model support first.** Confirm a quality-sufficient model is\n   available. No supported model meeting your bar → no engagement, regardless\n   of speed.\n2. **Benchmark head-to-head.** Same eval set, same prompts: Cerebras vs. your\n   current provider (and vs. other speed specialists). Measure quality, TTFT,\n   tokens\u002Fsec, and cost per task.\n3. **Test your context lengths.** Wafer-scale advantages can vary with sequence\n   length. Benchmark at your real context sizes, including your p99.\n4. **Validate output behavior.** Run your full eval suite — confirm the\n   hardware-optimized serving preserves the quality and behavior you need.\n5. **Model the economics.** Per-token price × tokens per task ÷ success rate =\n   cost per completed task. Compare honestly against slower, cheaper options.\n6. **Plan capacity and fallbacks.** Map rate limits to production traffic;\n   configure fallback providers for resilience.\n7. **Monitor production performance.** Track latency percentiles and quality\n   signals on live traffic; alert on drift.\n\nChecklist for Cerebras in production:\n- Supported model validated on your eval set.\n- Head-to-head benchmark vs. alternatives (quality + speed + cost).\n- Context-length behavior tested at your p99.\n- Rate limits mapped to production needs; fallbacks configured.\n- Cost per completed task modeled and acceptable.\n\n## Common pitfalls\n\n- **Speed as the only criterion.** Ultra-fast inference of inadequate quality\n  is worthless. Quality bar first, always.\n- **Headline-number anchoring.** Marketing tokens\u002Fsec measured under ideal\n  conditions. Your prompts, your context lengths, your concurrency — measure\n  those.\n- **Model availability assumptions.** Building plans around a model before\n  confirming it's supported. Check first.\n- **No head-to-head evaluation.** Choosing on reputation rather than measured\n  comparison. The speed-specialist field is competitive — benchmark.\n- **Ignoring the cost per task.** Faster per-token pricing can still lose on\n  cost per completed task if it uses more tokens or fails more often.\n- **Single-provider fragility.** Production on one speed specialist without\n  fallbacks. Fast outages are still outages.\n- **Context-length blind spots.** Validating speed at 1k tokens, deploying at\n  32k. Performance characteristics change with length — test yours.\n- **Skipping re-validation on updates.** Supported model versions change;\n  behavior can shift. Re-run evals when versions move.\n",{"name":5,"category":24,"description":7},"ai-research","main","af388a77c2fe5e0b4fd879cca4300a0ffa9a6579","skills\u002Fams-cerebras-guide",1,"https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis","https:\u002F\u002Fgithub.com\u002FTanNhatCMS","https:\u002F\u002Fmrkiet.name.vn","2026-11-07 14:23:08",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add TanNhatCMS\u002Fjarvis --skill cerebras-guide",{"tool":39,"command":40},"bunx","bunx skills add TanNhatCMS\u002Fjarvis --skill cerebras-guide",{"tool":42,"command":43},"pnpm","pnpm dlx skills add TanNhatCMS\u002Fjarvis --skill cerebras-guide","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fams-cerebras-guide\u002FSKILL.md",true,[],[48,55,63,70,77,84],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M7EWYJR0J8DARY7E59Z6Y9MV","web-scrape","Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open\u002Ffetch fails on JS-heavy pages, or when you need clean markdown\u002Fstructured extraction from a site. Requires FIRECRAWL_API_KEY.","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fweb-scrape\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":59,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M7EWYBN5P6F2AJG3FR24T9BP","design-taste-frontend","tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML\u002FCSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, \"AI-made\", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Ftaste-design\u002FLICENSE","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Ftaste-design\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M7EWY8FQFNQ2GB7NZZWGSSDX","test-driven-development","Use when implementing any feature or bugfix, before writing implementation code","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-tdd\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7EWY5SHBFRH1HDHVWM9K7TV","systematic-debugging","Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-debugging\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7EWY3NM0Z0JAKS7HDZ24K66","requesting-code-review","Use when completing tasks, implementing major features, or before merging to verify work meets requirements","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-code-review\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7EWY0WRMWX4M9ZFY0WKC593","brainstorming","You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-brainstorming\u002FSKILL.md",[],[]]