--- title: Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is... description: Kimi K3 tops the Next.js AI model leaderboard with a 92% success rate on Next.js-specific coding tasks, edging out Claude Sonnet 4.5 and Codex. However, the... canonical: https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is Better Than Claude Code / Codex? | daily.dev og:description: Kimi K3 tops the Next.js AI model leaderboard with a 92% success rate on Next.js-specific coding tasks, edging out Claude Sonnet 4.5 and Codex. However, the... og:url: https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1 og:image: https://api.daily.dev/og/posts/54Ups6zM1.png og:image:alt: Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is Better Than Claude Code / Codex? og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is Better Than Claude Code / Codex? **[ByteGrad](https://daily.dev/sources/bytegrad)** · 19 min read · 1 upvotes · 0 comments ## Summary Kimi K3 tops the Next.js AI model leaderboard with a 92% success rate on Next.js-specific coding tasks, edging out Claude Sonnet 4.5 and Codex. However, the more important finding is that bundling documentation directly into the project (via agents.md in Next.js 16.2+) boosts success rates significantly across all models — even older ones reach near-frontier performance. The video also covers Supabase's new agent skills, which bundle critical security rules (like RLS policies and service key exposure) directly into the project so AI coding agents can't miss them. A practical walkthrough shows how to set up OpenCode with Kimi K3 via the Vercel AI gateway and install Supabase agent skills. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Which AI model currently performs best on Next.js code generation benchmarks? Kimi K3, an open-weight model, currently ranks first on Vercel's Next.js-specific evals and on the arena.ai front-end code arena, surpassing Claude Fable 5 and other proprietary frontier models. It achieved a 92% success rate on Next.js code generation and migration tasks, missing only the proxy middleware rename and instant navigation tasks, and edged out Fable 5 mainly due to faster execution time. _Developers picking a coding agent for Next.js work can track these leaderboard shifts on daily.dev before locking in a model._ ### Does including agents.md documentation actually improve AI coding agent accuracy on Next.js tasks? Yes, bundling the agents.md file (which points agents to Next.js documentation stored in node_modules) raises success rates significantly, pushing Kimi K3 to 96% and letting older or mid-tier models like Claude 4.6 reach roughly the same success rate as frontier models. Since Next.js 16.2, this documentation is bundled directly into new projects, reducing dependence on which model or agent is used. _Teams choosing between frontier and cheaper models for Next.js can weigh this documentation-bundling effect on daily.dev._ ### What mistakes do AI coding agents commonly make with Supabase row-level security (RLS)? Coding agents frequently skip RLS policies on exposed schemas, hallucinate nonexistent CLI commands like 'supabase db execute', create views with security_invoker set to true (which silently bypasses row-level security), and sometimes ignore documentation entirely by relying on outdated training data. Supabase's open-source agent skills bundle a roughly 100-line skill.md with these security requirements to prevent such errors. _Developers wiring Supabase security into AI-assisted workflows can follow these agent pitfalls via daily.dev._ ## Similar posts on daily.dev - [I Tried Kimi K3 for a Week. Here’s What Happened.](https://daily.dev/posts/i-tried-kimi-k3-for-a-week-here-s-what-happened--ufveuen98) · Medium · 2 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#nextjs](https://daily.dev/tags/nextjs), [#supabase](https://daily.dev/tags/supabase), [#kimi](https://daily.dev/tags/kimi) [View this post on daily.dev](https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is Better Than Claude Code / Codex?","url":"https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1"},"datePublished":"2026-07-20T16:16:50.659Z","dateModified":"2026-09-14T06:27:30.036Z","description":"Kimi K3 tops the Next.js AI model leaderboard with a 92% success rate on Next.js-specific coding tasks, edging out Claude Sonnet 4.5 and Codex. However, the...","image":"https://i.ytimg.com/vi/zxQKxZqYLjc/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/zxQKxZqYLjc/sddefault.jpg","isAccessibleForFree":true,"articleSection":"ByteGrad","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"ByteGrad","logo":"https://media.daily.dev/image/upload/s--UnJ6vKLR--/f_auto/v1732552127/logos/bytegrad","url":"https://daily.dev/sources/bytegrad"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,nextjs,supabase,kimi","timeRequired":"PT19M","video":{"@type":"VideoObject","name":"Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is Better Than Claude Code / Codex?","description":"Kimi K3 tops the Next.js AI model leaderboard with a 92% success rate on Next.js-specific coding tasks, edging out Claude Sonnet 4.5 and Codex. However, the...","thumbnailUrl":"https://i.ytimg.com/vi/zxQKxZqYLjc/sddefault.jpg","uploadDate":"2026-07-20T16:16:50.659Z","duration":"PT19M","url":"https://api.daily.dev/r/54Ups6zM1","embedUrl":"https://www.youtube.com/embed/zxQKxZqYLjc"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"ByteGrad","item":"https://daily.dev/sources/bytegrad"},{"@type":"ListItem","position":3,"name":"Next.js AI-Model Leaderboard: Kimi K3 + OpenCode Is Better Than Claude Code / Codex?"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/next-js-ai-model-leaderboard-kimi-k3-opencode-is-better-than-claude-code-codex--54ups6zm1#faq","mainEntity":[{"@type":"Question","name":"Which AI model currently performs best on Next.js code generation benchmarks?","acceptedAnswer":{"@type":"Answer","text":"Kimi K3, an open-weight model, currently ranks first on Vercel's Next.js-specific evals and on the arena.ai front-end code arena, surpassing Claude Fable 5 and other proprietary frontier models. It achieved a 92% success rate on Next.js code generation and migration tasks, missing only the proxy middleware rename and instant navigation tasks, and edged out Fable 5 mainly due to faster execution time. Developers picking a coding agent for Next.js work can track these leaderboard shifts on daily.dev before locking in a model."}},{"@type":"Question","name":"Does including agents.md documentation actually improve AI coding agent accuracy on Next.js tasks?","acceptedAnswer":{"@type":"Answer","text":"Yes, bundling the agents.md file (which points agents to Next.js documentation stored in node_modules) raises success rates significantly, pushing Kimi K3 to 96% and letting older or mid-tier models like Claude 4.6 reach roughly the same success rate as frontier models. Since Next.js 16.2, this documentation is bundled directly into new projects, reducing dependence on which model or agent is used. Teams choosing between frontier and cheaper models for Next.js can weigh this documentation-bundling effect on daily.dev."}},{"@type":"Question","name":"What mistakes do AI coding agents commonly make with Supabase row-level security (RLS)?","acceptedAnswer":{"@type":"Answer","text":"Coding agents frequently skip RLS policies on exposed schemas, hallucinate nonexistent CLI commands like 'supabase db execute', create views with security_invoker set to true (which silently bypasses row-level security), and sometimes ignore documentation entirely by relying on outdated training data. Supabase's open-source agent skills bundle a roughly 100-line skill.md with these security requirements to prevent such errors. Developers wiring Supabase security into AI-assisted workflows can follow these agent pitfalls via daily.dev."}}]} ```