{"id":163,"date":"2026-01-20T00:00:00","date_gmt":"2026-01-19T23:00:00","guid":{"rendered":"https:\/\/helloblog.io\/sl\/wp-bench-uradni-benchmark-za-wordpress-ai\/"},"modified":"2026-01-20T00:00:00","modified_gmt":"2026-01-19T23:00:00","slug":"wp-bench-uradni-benchmark-za-wordpress-ai","status":"publish","type":"post","link":"https:\/\/helloblog.io\/sl\/wp-bench-uradni-benchmark-za-wordpress-ai\/","title":{"rendered":"WP-Bench: uradni benchmark za merjenje, kako dobro AI razume WordPress"},"content":{"rendered":"\n<p>Generi\u010dni programerski benchmarki pogosto poka\u017eejo, da modeli zlahka napi\u0161ejo splo\u0161no PHP ali JavaScript kodo. Pri WordPressu pa se zatakne na tipi\u010dnih stvareh: pravilna raba hookov (actions\/filters), varnostne prakse (nonces, capability checks), WordPress coding standards, arhitektura vti\u010dnikov in predvsem to, ali se koda dejansko obna\u0161a pravilno v pravem runtime okolju.<\/p>\n\n\n\n<p>Zato je WordPress projekt predstavil <strong>WP-Bench<\/strong> \u2013 uradni WordPress AI benchmark: odprtokodno orodje in nabor testov, ki merijo, kako dobro posamezni jezikovni modeli razumejo WordPress razvoj in kako dobro znajo generirati kodo, ki prestane realno izvajanje.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Kaj je WP-Bench in kaj pravzaprav meri<\/h2>\n\n\n\n<p>WP-Bench je benchmark (standardiziran testni okvir), ki AI modelom postavlja WordPress-specifi\u010dne naloge in jih ocenjuje v dveh dimenzijah:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n\n<li><strong>Knowledge<\/strong>: vpra\u0161anja tipa multiple-choice o WordPress konceptih, API-jih, hookih, varnostnih vzorcih in coding standardih. Poudarek je tudi na sodobnej\u0161ih dodatkih, kot sta <em>Abilities API<\/em> in <em>Interactivity API<\/em>.<\/li>\n\n\n<li><strong>Execution<\/strong>: naloge generiranja kode, ki se ocenijo prek pravega WordPress runtime-a \u2013 z analizo kode in z dejanskim izvajanjem ter assertions (trditvami\/testi), ki preverijo rezultat.<\/li>\n\n<\/ul>\n\n\n\n<p>Klju\u010dna ideja: model lahko \u201cpozna teorijo\u201d, a pade na prakti\u010dnih detajlih (npr. napa\u010den hook, manjkajo\u010d nonce, napa\u010dna sanitizacija, neupo\u0161tevanje standardov). WP-Bench \u017eeli meriti oboje \u2013 razumevanje in izvedbo.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Zakaj je to pomembno za WordPress razvijalce<\/h2>\n\n\n\n<p>WordPress poganja ogromen del spleta, a ve\u010dina javnih ocen AI modelov temelji na splo\u0161nih nalogah. To je problemati\u010dno, ker WordPress razvoj ni samo \u201cnapi\u0161i funkcijo\u201d, ampak specifi\u010den ekosistem: hooki, WP_Query, REST API, permissions, multisite posebnosti, varnost in kompatibilnost.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n\n<li><strong>Bolj informirana izbira orodij<\/strong>: \u010de gradi\u0161 AI-podprte funkcionalnosti v vti\u010dniku ali pa uporablja\u0161 coding assistant, je koristno vedeti, kateri modeli se v WordPress kontekstu dejansko izka\u017eejo.<\/li>\n\n\n<li><strong>Pritisk na ponudnike modelov<\/strong>: cilj je, da WP-Bench postane standard, ki ga laboratoriji in ponudniki modelov vklju\u010dijo v predizdajne evaluacije \u2013 tako WordPress zmogljivost ne bo \u201cnaknadna misel\u201d.<\/li>\n\n\n<li><strong>Transparentnost prek leaderboarda<\/strong>: ekipa dela proti javnemu leaderboardu, kjer bi bili rezultati primerljivi, ponovljivi in uporabni za skupnost.<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Kako poteka ocenjevanje (grading) v praksi<\/h2>\n\n\n\n<p>WP-Bench ne ocenjuje samo z \u201cheuristiko\u201d ali subjektivnim pregledom. Namesto tega uporabi WordPress kot ocenjevalca: generirano kodo po\u017eene v izoliranem okolju in jo preveri s stati\u010dno analizo ter runtime testi.<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n\n<li>Harness po\u0161lje modelu prompt z zahtevo, da napi\u0161e WordPress kodo.<\/li>\n\n\n<li>Generirana koda se posreduje WordPress runtime-u prek WP-CLI.<\/li>\n\n\n<li>Runtime izvede stati\u010dno analizo (npr. sintaksa, coding standards, varnostni vzorci).<\/li>\n\n\n<li>Koda se za\u017eene v sandboxu, kjer assertions preverijo pri\u010dakovano obna\u0161anje.<\/li>\n\n\n<li>Rezultati se vrnejo kot JSON s to\u010dkami in podrobnim feedbackom.<\/li>\n\n<\/ol>\n\n\n\n<p>Ta pristop je zanimiv predvsem zato, ker odpre vrata za ponovljive primerjave: isti testni set, enak runtime, enaka pravila ocenjevanja. \u010ce model \u201chalucinira\u201d WordPress API ali naredi tipi\u010dno napako, se to hitro poka\u017ee.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Hiter zagon: od namestitve do prvih rezultatov<\/h2>\n\n\n\n<p>Repo je odprtokoden in zasnovan tako, da lahko benchmark po\u017eene\u0161 lokalno. Minimalni potek izgleda takole.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1) Namesti Python del (harness)<\/h3>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>python3 -m venv .venv &amp;&amp; source .venv\/bin\/activate\npip install -e .\/python\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#B392F0\">python3<\/span><span style=\"color:#79B8FF\"> -m<\/span><span style=\"color:#9ECBFF\"> venv<\/span><span style=\"color:#9ECBFF\"> .venv<\/span><span style=\"color:#E1E4E8\"> &#x26;&#x26; <\/span><span style=\"color:#79B8FF\">source<\/span><span style=\"color:#9ECBFF\"> .venv\/bin\/activate<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">pip<\/span><span style=\"color:#9ECBFF\"> install<\/span><span style=\"color:#79B8FF\"> -e<\/span><span style=\"color:#9ECBFF\"> .\/python<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<h3 class=\"wp-block-heading\">2) Nastavi API klju\u010de ponudnikov modelov<\/h3>\n\n\n\n<p>V <code>.env<\/code> doda\u0161 klju\u010de za tiste ponudnike, ki jih \u017eeli\u0161 testirati (npr. OpenAI, Anthropic, Google):<\/p>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>OPENAI_API_KEY=sk-...\nANTHROPIC_API_KEY=sk-ant-...\nGOOGLE_API_KEY=...\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#E1E4E8\">OPENAI_API_KEY<\/span><span style=\"color:#F97583\">=<\/span><span style=\"color:#9ECBFF\">sk-...<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">ANTHROPIC_API_KEY<\/span><span style=\"color:#F97583\">=<\/span><span style=\"color:#9ECBFF\">sk-ant-...<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">GOOGLE_API_KEY<\/span><span style=\"color:#F97583\">=<\/span><span style=\"color:#9ECBFF\">...<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<h3 class=\"wp-block-heading\">3) Za\u017eeni WordPress runtime (grader)<\/h3>\n\n\n\n<p>Runtime del v repozitoriju uporablja Node\/NPM in vsebuje WordPress grader plugin ter konfiguracijo za okolje.<\/p>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>cd runtime\nnpm install\nnpm start\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#79B8FF\">cd<\/span><span style=\"color:#9ECBFF\"> runtime<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">npm<\/span><span style=\"color:#9ECBFF\"> install<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">npm<\/span><span style=\"color:#9ECBFF\"> start<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<h3 class=\"wp-block-heading\">4) Po\u017eeni benchmark<\/h3>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>cd ..\nwp-bench run --config wp-bench.example.yaml\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#79B8FF\">cd<\/span><span style=\"color:#9ECBFF\"> ..<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">wp-bench<\/span><span style=\"color:#9ECBFF\"> run<\/span><span style=\"color:#79B8FF\"> --config<\/span><span style=\"color:#9ECBFF\"> wp-bench.example.yaml<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<p>Rezultati se zapi\u0161ejo v <code>output\/results.json<\/code>, dodatni per-test logi pa v <code>output\/results.jsonl<\/code>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Primerjava ve\u010d modelov v enem zagonu<\/h2>\n\n\n\n<p>WP-Bench podpira multi-model zagon: v konfiguracijo na\u0161teje\u0161 ve\u010d modelov, harness pa jih po\u017eene zaporedno in na koncu izpi\u0161e primerjalno tabelo. Imena modelov sledijo konvencijam LiteLLM.<\/p>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>models:\n  - name: gpt-4o\n  - name: gpt-4o-mini\n  - name: claude-sonnet-4-20250514\n  - name: claude-opus-4-5-20251101\n  - name: gemini\/gemini-2.5-pro\n  - name: gemini\/gemini-2.5-flash\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#85E89D\">models<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">gpt-4o<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">gpt-4o-mini<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">claude-sonnet-4-20250514<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">claude-opus-4-5-20251101<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">gemini\/gemini-2.5-pro<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">gemini\/gemini-2.5-flash<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<p>To je uporabno, ko \u017eeli\u0161 z lastnimi testi preveriti, kateri model je najbolj\u0161i za tipi\u010dne naloge v tvojem workflowu (npr. pisanje vti\u010dnikov, refaktoriranje, generiranje REST endpointov) \u2013 in to v okolju, kjer se napake poka\u017eejo kot failed assertions, ne kot \u201clep\u201d odgovor.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Konfiguracija: dataset, grader in parametri zagona<\/h2>\n\n\n\n<p>Osnovni na\u010din je, da kopira\u0161 <code>wp-bench.example.yaml<\/code> in ga prilagodi\u0161. Tipi\u010dno nastavi\u0161 izvor dataseta, model(e), na\u010din ocenjevanja ter omejitve (limit) in paralelizem (concurrency).<\/p>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>dataset:\n  source: local              # 'local' or 'huggingface'\n  name: wp-core-v1           # suite name\n\nmodels:\n  - name: gpt-4o\n\ngrader:\n  kind: docker\n  wp_env_dir: .\/runtime      # path to wp-env project\n\nrun:\n  suite: wp-core-v1\n  limit: 10                  # limit tests (null = all)\n  concurrency: 4\n\noutput:\n  path: output\/results.json\n  jsonl_path: output\/results.jsonl\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#85E89D\">dataset<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  source<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">local<\/span><span style=\"color:#6A737D\">              # 'local' or 'huggingface'<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">wp-core-v1<\/span><span style=\"color:#6A737D\">           # suite name<\/span><\/span>\n<span class=\"line\"><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">models<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#E1E4E8\">  - <\/span><span style=\"color:#85E89D\">name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">gpt-4o<\/span><\/span>\n<span class=\"line\"><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">grader<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  kind<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">docker<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  wp_env_dir<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">.\/runtime<\/span><span style=\"color:#6A737D\">      # path to wp-env project<\/span><\/span>\n<span class=\"line\"><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">run<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  suite<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">wp-core-v1<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  limit<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#79B8FF\">10<\/span><span style=\"color:#6A737D\">                  # limit tests (null = all)<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  concurrency<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#79B8FF\">4<\/span><\/span>\n<span class=\"line\"><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">output<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  path<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">output\/results.json<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  jsonl_path<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">output\/results.jsonl<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<h3 class=\"wp-block-heading\">Uporabni CLI ukazi<\/h3>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>wp-bench run --config wp-bench.yaml          # run with config file\nwp-bench run --model-name gpt-4o --limit 5   # quick single-model test\nwp-bench dry-run --config wp-bench.yaml      # validate config without calling models\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#B392F0\">wp-bench<\/span><span style=\"color:#9ECBFF\"> run<\/span><span style=\"color:#79B8FF\"> --config<\/span><span style=\"color:#9ECBFF\"> wp-bench.yaml<\/span><span style=\"color:#6A737D\">          # run with config file<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">wp-bench<\/span><span style=\"color:#9ECBFF\"> run<\/span><span style=\"color:#79B8FF\"> --model-name<\/span><span style=\"color:#9ECBFF\"> gpt-4o<\/span><span style=\"color:#79B8FF\"> --limit<\/span><span style=\"color:#79B8FF\"> 5<\/span><span style=\"color:#6A737D\">   # quick single-model test<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">wp-bench<\/span><span style=\"color:#9ECBFF\"> dry-run<\/span><span style=\"color:#79B8FF\"> --config<\/span><span style=\"color:#9ECBFF\"> wp-bench.yaml<\/span><span style=\"color:#6A737D\">      # validate config without calling models<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<h2 class=\"wp-block-heading\">Struktura repozitorija in kje \u017eivijo testi<\/h2>\n\n\n\n<p>Repo je precej jasno razdeljen: Python del vodi izvedbo (harness), runtime del skrbi za WordPress grader, dataseti pa so organizirani v suites.<\/p>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>.\n\u251c\u2500\u2500 python\/          # Benchmark harness (pip installable)\n\u251c\u2500\u2500 runtime\/         # WordPress grader plugin + wp-env config\n\u251c\u2500\u2500 datasets\/        # Test suites (local JSON + Hugging Face builder)\n\u251c\u2500\u2500 notebooks\/       # Results visualization and reporting\n\u2514\u2500\u2500 output\/          # Benchmark results (gitignored)\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#79B8FF\">.<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">\u251c\u2500\u2500<\/span><span style=\"color:#9ECBFF\"> python\/<\/span><span style=\"color:#6A737D\">          # Benchmark harness (pip installable)<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">\u251c\u2500\u2500<\/span><span style=\"color:#9ECBFF\"> runtime\/<\/span><span style=\"color:#6A737D\">         # WordPress grader plugin + wp-env config<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">\u251c\u2500\u2500<\/span><span style=\"color:#9ECBFF\"> datasets\/<\/span><span style=\"color:#6A737D\">        # Test suites (local JSON + Hugging Face builder)<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">\u251c\u2500\u2500<\/span><span style=\"color:#9ECBFF\"> notebooks\/<\/span><span style=\"color:#6A737D\">       # Results visualization and reporting<\/span><\/span>\n<span class=\"line\"><span style=\"color:#B392F0\">\u2514\u2500\u2500<\/span><span style=\"color:#9ECBFF\"> output\/<\/span><span style=\"color:#6A737D\">          # Benchmark results (gitignored)<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<p>Test suite-i so v <code>datasets\/suites\/&lt;suite-name&gt;\/<\/code> in imajo tipi\u010dno dva sklopa:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n\n<li><code>execution\/<\/code> \u2013 naloge generiranja kode z assertions (po kategorijah, v JSON datotekah)<\/li>\n\n\n<li><code>knowledge\/<\/code> \u2013 multiple-choice vpra\u0161anja (prav tako po kategorijah, v JSON datotekah)<\/li>\n\n<\/ul>\n\n\n\n<p>Privzeti suite <code>wp-core-v1<\/code> pokriva WordPress core API-je, hooks, delo z bazo in varnostne vzorce.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Nalagalnik dataseta prek Hugging Face<\/h3>\n\n\n\n<p>\u010ce \u017eeli\u0161 dataset uporabljati iz Hugging Face vira, konfiguracija izgleda takole:<\/p>\n\n\n\n<div class=\"wp-block-kevinbatdorf-code-block-pro\" data-code-block-pro-font-family=\"Code-Pro-JetBrains-Mono\" style=\"font-size:.875rem;font-family:Code-Pro-JetBrains-Mono,ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace;line-height:1.25rem;--cbp-tab-width:2;tab-size:var(--cbp-tab-width, 2)\"><span style=\"display:block;padding:16px 0 0 16px;margin-bottom:-1px;width:100%;text-align:left;background-color:#24292e\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"54\" height=\"14\" viewBox=\"0 0 54 14\"><g fill=\"none\" fill-rule=\"evenodd\" transform=\"translate(1 1)\"><circle cx=\"6\" cy=\"6\" r=\"6\" fill=\"#FF5F56\" stroke=\"#E0443E\" stroke-width=\".5\"><\/circle><circle cx=\"26\" cy=\"6\" r=\"6\" fill=\"#FFBD2E\" stroke=\"#DEA123\" stroke-width=\".5\"><\/circle><circle cx=\"46\" cy=\"6\" r=\"6\" fill=\"#27C93F\" stroke=\"#1AAB29\" stroke-width=\".5\"><\/circle><\/g><\/svg><\/span><span role=\"button\" tabindex=\"0\" style=\"color:#e1e4e8;display:none\" aria-label=\"Copy\" class=\"code-block-pro-copy-button\"><pre class=\"code-block-pro-copy-button-pre\" aria-hidden=\"true\"><textarea class=\"code-block-pro-copy-button-textarea\" tabindex=\"-1\" aria-hidden=\"true\" readonly>dataset:\n  source: huggingface\n  name: WordPress\/wp-bench-v1\n<\/textarea><\/pre><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"width:24px;height:24px\" fill=\"none\" viewBox=\"0 0 24 24\" stroke=\"currentColor\" stroke-width=\"2\"><path class=\"with-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2m-6 9l2 2 4-4\"><\/path><path class=\"without-check\" stroke-linecap=\"round\" stroke-linejoin=\"round\" d=\"M9 5H7a2 2 0 00-2 2v12a2 2 0 002 2h10a2 2 0 002-2V7a2 2 0 00-2-2h-2M9 5a2 2 0 002 2h2a2 2 0 002-2M9 5a2 2 0 012-2h2a2 2 0 012 2\"><\/path><\/svg><\/span><pre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\"><code><span class=\"line\"><span style=\"color:#85E89D\">dataset<\/span><span style=\"color:#E1E4E8\">:<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  source<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">huggingface<\/span><\/span>\n<span class=\"line\"><span style=\"color:#85E89D\">  name<\/span><span style=\"color:#E1E4E8\">: <\/span><span style=\"color:#9ECBFF\">WordPress\/wp-bench-v1<\/span><\/span><\/code><\/pre><\/div>\n\n\n\n<h2 class=\"wp-block-heading\">Trenutno stanje in omejitve, ki jih je dobro poznati<\/h2>\n\n\n\n<p>WP-Bench je zgodnja izdaja, zato je ekipa precej jasna glede omejitev:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n\n<li><strong>Velikost dataseta<\/strong>: trenutni testni set je razmeroma majhen; za res celovito sliko bo potrebnih ve\u010d primerov \u010dez \u0161ir\u0161i spekter WordPress API-jev in tipi\u010dnih razvojnih vzorcev.<\/li>\n\n\n<li><strong>Pristranskost k novej\u0161im verzijam<\/strong>: suite trenutno nagiba k funkcionalnostim okoli WordPress 6.9 (npr. Abilities API, Interactivity API). To je delno namenoma, ker se ravno pri novih API-jih modeli pogosto najbolj lovijo, a hkrati to lahko vpliva na primerljivost z modeli, treniranimi na starej\u0161ih podatkih.<\/li>\n\n\n<li><strong>Saturacija pri starej\u0161ih konceptih<\/strong>: pri zgodnjih testih so modeli dosegali zelo visoke rezultate na starej\u0161ih WordPress temah, kar pomeni, da tak\u0161na vpra\u0161anja ne dajejo ve\u010d dobrega signala. Izziv je najti naloge, ki so res te\u017eke \u2013 ne zgolj nove.<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Kdaj ti WP-Bench realno pride prav<\/h2>\n\n\n\n<p>\u010ce v ekipi razmi\u0161lja\u0161 o uvedbi AI v razvojni proces, je najve\u010dja vrednost WP-Bench v tem, da omogo\u010da <strong>ponovljiv preizkus<\/strong> za WordPress kontekst. Na primer:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n\n<li>izbira modela za interni coding assistant, ko ti je pomembna skladnost s WordPress coding standards in varnostnimi praksami<\/li>\n\n\n<li>primerjava modelov pred tem, ko v produkt vgradi\u0161 AI-generirano kodo (npr. \u201csite builder\u201d funkcije, generator blokov, scaffolding za vti\u010dnike)<\/li>\n\n\n<li>regresijsko testiranje: ko menja\u0161 model ali ponudnika, \u017eeli\u0161 videti, ali kakovost WordPress-specifi\u010dne kode pade ali zraste<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Viri<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n\n<li>Repo: <a href=\"https:\/\/github.com\/WordPress\/wp-bench\">WP-Bench<\/a><\/li>\n\n\n<li>Kontekst: <a href=\"https:\/\/make.wordpress.org\/ai\/2025\/07\/17\/ai-building-blocks\/\">AI Building Blocks for WordPress<\/a><\/li>\n\n\n<li>WordPress Slack kanal: <a href=\"https:\/\/wordpress.slack.com\/archives\/C08TJ8BPULS\">#core-ai<\/a><\/li>\n\n<\/ul>\n\n\n<div class=\"references-section\">\n                <h2>Reference \/ Viri<\/h2>\n                <ul class=\"references-list\"><li><a href=\"https:\/\/make.wordpress.org\/ai\/2026\/01\/14\/introducing-wp-bench-a-wordpress-ai-benchmark\/\" target=\"_blank\" rel=\"noopener noreferrer\">Introducing WP-Bench: A WordPress AI Benchmark<\/a><\/li><li><a href=\"https:\/\/github.com\/WordPress\/wp-bench\" target=\"_blank\" rel=\"noopener noreferrer\">WP-Bench GitHub README<\/a><\/li><\/ul>\n            <\/div>","protected":false},"excerpt":{"rendered":"<p>\u010ce uporablja\u0161 AI pri pisanju WordPress kode, hitro ugotovi\u0161, da \u201cdobro programira\u201d \u0161e ne pomeni \u201cpozna WordPress\u201d. WP-Bench je nov uradni benchmark, ki to razliko kon\u010dno izmeri na sistemati\u010den na\u010din.<\/p>\n","protected":false},"author":45,"featured_media":162,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[17],"tags":[28,80,10,8,81],"class_list":["post-163","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ekosistem-wordpress","tag-ai","tag-benchmark","tag-wordpress","tag-wp-cli","tag-wp-env"],"_links":{"self":[{"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/posts\/163","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/users\/45"}],"replies":[{"embeddable":true,"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/comments?post=163"}],"version-history":[{"count":0,"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/posts\/163\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/media\/162"}],"wp:attachment":[{"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/media?parent=163"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/categories?post=163"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/helloblog.io\/sl\/wp-json\/wp\/v2\/tags?post=163"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}