{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-09-28T06:03:00.468Z","headline":"我在同一项 Three.js 任务上测试了 10 种模型/连接线组合","description":"一项非正式测试用同一段 Three.js 科幻机库构建提示词，对比 10 种模型与连接线（harness）组合的代码生成表现。GLM 5.3 Flash Max 搭配 OpenCode 用时最短（20 分钟）且得分最高（96.89%），Astra 6.0 Max 搭配 Codex 耗时最长（37 分钟）。测试记录了各组合的耗时、token 用量与工具调用成功率。","url":"https://www.aioga.com/news/cmtspf1bu024xroou4hxsgbwc/","mainEntityOfPage":"https://www.aioga.com/news/cmtspf1bu024xroou4hxsgbwc/","datePublished":"2026-09-08T13:07:16.000Z","dateModified":"2026-09-08T13:07:16.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://alvins82.github.io/hangar-harness-model-tests","https://aihot.news/items/cmtspf1bu024xroou4hxsgbwc"],"canonicalUrl":"https://www.aioga.com/news/cmtspf1bu024xroou4hxsgbwc/","directAnswer":{"@type":"Answer","text":"材料介绍了一项非正式 Three.js 代码生成测试：研究者使用同一段科幻机库构建提示词，对比 10 种模型与连接线组合，并记录耗时、token 用量和工具调用成功率。","url":"https://www.aioga.com/news/cmtspf1bu024xroou4hxsgbwc/","dateCreated":"2026-09-08T13:07:16.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"alvins82.github.io source article","url":"https://alvins82.github.io/hangar-harness-model-tests","datePublished":"2026-09-08T13:07:16.000Z","provider":{"@type":"Organization","name":"alvins82.github.io","url":"https://alvins82.github.io/hangar-harness-model-tests"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/cmtspf1bu024xroou4hxsgbwc","datePublished":"2026-09-08T13:07:16.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/cmtspf1bu024xroou4hxsgbwc"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"alvins82.github.io","url":"https://alvins82.github.io/hangar-harness-model-tests"},"geoDeepAnswer":null,"article":{"id":"cmtspf1bu024xroou4hxsgbwc","slug":"cmtspf1bu024xroou4hxsgbwc","url":"https://www.aioga.com/news/cmtspf1bu024xroou4hxsgbwc/","title":"我在同一项 Three.js 任务上测试了 10 种模型/连接线组合","title_en":"","summary":"一项非正式测试用同一段 Three.js 科幻机库构建提示词，对比 10 种模型与连接线（harness）组合的代码生成表现。GLM 5.3 Flash Max 搭配 OpenCode 用时最短（20 分钟）且得分最高（96.89%），Astra 6.0 Max 搭配 Codex 耗时最长（37 分钟）。测试记录了各组合的耗时、token 用量与工具调用成功率。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://alvins82.github.io/hangar-harness-model-tests","aiHotUrl":"https://aihot.news/items/cmtspf1bu024xroou4hxsgbwc","publishedAt":"2026-09-08T13:07:16.000Z","category":"行业动态","score":58,"selected":false,"articleBody":["I've been testing a simple prompt with different model and harness combinations to work out which one produces best results. I do this in /goal mode.","Prompt: Build a single-page Three.js sci-fi hangar with hovering drones, animated warning lights, emissive runway strips, and subtle volumetric-style fog planes. Include drone formation toggle and cinematic camera path. Output one self-contained HTML file with inline JavaScript.","All GLM runs are labelled GLM 5.3 Flash Max ; input tokens include cached input. Output tokens are the generated total, including reasoning; when a harness reports reasoning separately, the reasoning column shows that subset. The DSH adapter does not report a separate reasoning count. DSH durations sum active turn time across both turns, excluding the pause between turns. Tool errors are recorded failed tool events. A dash means unavailable or not reported."],"articleImages":[],"mediaStatus":"none","articleBodyZh":["我一直在使用不同模型和框架组合测试一个简单的提示，以找出哪一个能产生最佳效果。我在 /goal 模式下执行此操作。","提示：构建一个单页的 Three.js 科幻机库，包括悬浮无人机、动画警告灯、发光跑道条带和微妙的体积风格雾平面。包括无人机编队切换和电影感摄像机路径。输出一个自包含的 HTML 文件，并内联 JavaScript。","所有 GLM 运行都标记为 GLM 5.3 Flash Max；输入 token 包括缓存的输入。输出 token 是生成的总数，包括推理；当框架单独报告推理时，推理列显示该子集。DSH 适配器不报告单独的推理计数。DSH 持续时间是两次轮次的活动时间之和，不包括轮次之间的暂停。工具错误记录失败的工具事件。破折号表示不可用或未报告。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"材料介绍了一项非正式 Three.js 代码生成测试：研究者使用同一段科幻机库构建提示词，对比 10 种模型与连接线组合，并记录耗时、token 用量和工具调用成功率。","background":"测试要求生成单页 Three.js 科幻机库，包含悬浮无人机、动画警示灯、发光跑道、雾效平面、编队切换和电影感摄像机路径，并输出内嵌 JavaScript 的单文件 HTML。摘要称 GLM 5.3 Flash Max 搭配 OpenCode 用时最短且得分最高。","viewpoint":"Aioga 判断：这项测试能提供特定提示词和工作模式下的横向参考，但属于非正式对比，结果不足以单独证明某一模型或连接线在所有 Three.js 任务中都更优。","implications":"可能影响：模型评测不应只看最终得分，还需要结合耗时、token 统计和工具调用成功率；由于不同连接线的统计口径存在差异，相关结果可能需要谨慎比较，不代表普遍性能结论。","nextStep":"后续观察：可关注测试是否公开更完整的评分细则、各组合的原始输出和失败工具事件，并在更多提示词与任务类型上重复验证，以判断结果是否具有更广泛的参考价值。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-09-08T14:06:06.680Z","sourceHash":"139d10018cd8cc59","review":{"approved":true,"groundedness":94,"clarity":92,"duplicationRisk":20,"blockingIssues":[],"notes":["“Aioga 判断”已明确标注为观点，未冒充来源事实。","“结果可能需要谨慎比较”属于基于来源中统计口径差异的合理限定。","可在后续编辑中补充最短用时 20 分钟和最高得分 96.89%，但这属于可选的信息完整性优化，不影响审核通过。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","editorial-labels","inference-boundary","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"我在同一项 Three.js 任务上测试了 10 种模型/连接线组合","summary":"一项非正式测试用同一段 Three.js 科幻机库构建提示词，对比 10 种模型与连接线（harness）组合的代码生成表现。GLM 5.3 Flash Max 搭配 OpenCode 用时最短（20 分钟）且得分最高（96.89%），Astra 6.0 Max 搭配 Codex 耗时最长（37 分钟）。测试记录了各组合的耗时、token 用量与工具调用成功率。","category":"行业动态","source":"alvins82.github.io","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"我在同一项 Three.js 任务上测试了 10 种模型/连接线组合 - Aioga AI资讯","description":"一项非正式测试用同一段 Three.js 科幻机库构建提示词，对比 10 种模型与连接线（harness）组合的代码生成表现。GLM 5.3 Flash Max 搭配 OpenCode 用时最短（20 分钟）且得分最高（96.89%），Astra 6.0 Max 搭配 Codex 耗时最长（37 分钟）。测试记录了各组合的耗时、token 用量与工具调用成功...","url":"https://www.aioga.com/news/cmtspf1bu024xroou4hxsgbwc/","articleBody":["我一直在使用不同模型和框架组合测试一个简单的提示，以找出哪一个能产生最佳效果。我在 /goal 模式下执行此操作。","提示：构建一个单页的 Three.js 科幻机库，包括悬浮无人机、动画警告灯、发光跑道条带和微妙的体积风格雾平面。包括无人机编队切换和电影感摄像机路径。输出一个自包含的 HTML 文件，并内联 JavaScript。","所有 GLM 运行都标记为 GLM 5.3 Flash Max；输入 token 包括缓存的输入。输出 token 是生成的总数，包括推理；当框架单独报告推理时，推理列显示该子集。DSH 适配器不报告单独的推理计数。DSH 持续时间是两次轮次的活动时间之和，不包括轮次之间的暂停。工具错误记录失败的工具事件。破折号表示不可用或未报告。"]},"en":{"title":"I tested 10 model/connection combinations on the same Three.js task","summary":"An informal test used the same Three.js sci-fi hangar construction prompt to compare the code generation performance of 10 model and connection (harness) combinations. GLM 5.3 Flash Max paired with OpenCode took the shortest time (20 minutes) and achieved the highest score (96.89%), while Astra 6.0 Max paired with Codex took the longest (37 minutes). The test recorded the time, token usage, and tool call success rate for each combination.","category":"Industry","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"I tested 10 model/connection combinations on the same Three.js task - Aioga AI News","description":"An informal test used the same Three.js sci-fi hangar construction prompt to compare the code generation performance of 10 model and connection (harness) combinations. GLM 5.3 Flas...","url":"https://www.aioga.com/en/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:39.610Z"},"ja":{"title":"同じ Three.js タスクで 10 種類のモデル/接続線をテスト","summary":"非公式テストとして、同じ Three.js のサイエンスフィクション機体格納庫構築のプロンプトで、10 種類のモデルと接続線（harness）の組み合わせのコード生成パフォーマンスを比較。GLM 5.3 Flash Max と OpenCode の組み合わせは最短時間（20 分）で最も高いスコア（96.89%）を獲得、Astra 6.0 Max と Codex の組み合わせは最長時間（37 分）を要した。各組み合わせの時間、トークン消費量、およびツール呼び出し成功率も記録。","category":"業界動向","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"同じ Three.js タスクで 10 種類のモデル/接続線をテスト - Aioga AIニュース","description":"非公式テストとして、同じ Three.js のサイエンスフィクション機体格納庫構築のプロンプトで、10 種類のモデルと接続線（harness）の組み合わせのコード生成パフォーマンスを比較。GLM 5.3 Flash Max と OpenCode の組み合わせは最短時間（20 分）で最も高いスコア（96.89%）を獲得、Astra 6.0 Max と Code...","url":"https://www.aioga.com/ja/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:43.209Z"},"ko":{"title":"나는 동일한 Three.js 과제에서 10가지 모델/연결선 조합을 테스트했다","summary":"비공식 테스트에서 동일한 Three.js 공상 과학 기지 구축 프롬프트를 사용하여 10가지 모델과 연결선(harness) 조합의 코드 생성 성능을 비교했다. GLM 5.3 Flash Max와 OpenCode 조합이 가장 짧은 시간(20분)과 최고 점수(96.89%)를 기록했으며, Astra 6.0 Max와 Codex 조합은 가장 오래 걸렸다(37분). 테스트에서는 각 조합의 소요 시간, 토큰 사용량, 도구 호출 성공률을 기록했다.","category":"업계 동향","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"나는 동일한 Three.js 과제에서 10가지 모델/연결선 조합을 테스트했다 - Aioga AI 뉴스","description":"비공식 테스트에서 동일한 Three.js 공상 과학 기지 구축 프롬프트를 사용하여 10가지 모델과 연결선(harness) 조합의 코드 생성 성능을 비교했다. GLM 5.3 Flash Max와 OpenCode 조합이 가장 짧은 시간(20분)과 최고 점수(96.89%)를 기록했으며, Astra 6.0 Max와 Codex 조합...","url":"https://www.aioga.com/ko/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:47.920Z"},"es":{"title":"Probé 10 combinaciones de modelo/conexión en la misma tarea de Three.js","summary":"Una prueba informal utilizó el mismo conjunto de indicaciones de construcción para un hangar futurista de Three.js, comparando la generación de código de 10 combinaciones de modelos y conexiones (harness). GLM 5.3 Flash Max con OpenCode tardó menos tiempo (20 minutos) y obtuvo la puntuación más alta (96.89%), mientras que Astra 6.0 Max con Codex tardó más (37 minutos). La prueba registró el tiempo de ejecución, el uso de tokens y la tasa de éxito en llamadas a herramientas de cada combinación.","category":"Industria","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Probé 10 combinaciones de modelo/conexión en la misma tarea de Three.js - Aioga Noticias de IA","description":"Una prueba informal utilizó el mismo conjunto de indicaciones de construcción para un hangar futurista de Three.js, comparando la generación de código de 10 combinaciones de modelo...","url":"https://www.aioga.com/es/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:50.227Z"},"fr":{"title":"I tested 10 model/connection combinations on the same Three.js task","summary":"An informal test used the same Three.js sci-fi hangar building prompt to compare the code generation performance of 10 model and harness combinations. GLM 5.3 Flash Max paired with OpenCode took the shortest time (20 minutes) and scored the highest (96.89%), while Astra 6.0 Max paired with Codex took the longest (37 minutes). The test recorded each combination’s time, token usage, and tool invocation success rate.","category":"Industrie","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"I tested 10 model/connection combinations on the same Three.js task - Aioga Actualités IA","description":"An informal test used the same Three.js sci-fi hangar building prompt to compare the code generation performance of 10 model and harness combinations. GLM 5.3 Flash Max paired with...","url":"https://www.aioga.com/fr/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:54.170Z"},"de":{"title":"Ich habe 10 Modell-/Verbindungskombinationen mit derselben Three.js-Aufgabe getestet","summary":"Ein informeller Test verwendete denselben Prompt für den Aufbau eines Science-Fiction-Hangars mit Three.js, um die Codegenerierungsleistung von 10 Modell- und Verbindungskombinationen (Harness) zu vergleichen. GLM 5.3 Flash Max in Kombination mit OpenCode benötigte am kürzesten (20 Minuten) und erzielte die höchste Punktzahl (96,89%), während Astra 6.0 Max mit Codex am längsten dauerte (37 Minuten). Der Test protokollierte die benötigte Zeit, den Tokenverbrauch und die Erfolgsrate der Toolaufrufe für jede Kombination.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Ich habe 10 Modell-/Verbindungskombinationen mit derselben Three.js-Aufgabe getestet - Aioga KI-News","description":"Ein informeller Test verwendete denselben Prompt für den Aufbau eines Science-Fiction-Hangars mit Three.js, um die Codegenerierungsleistung von 10 Modell- und Verbindungskombinatio...","url":"https://www.aioga.com/de/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:54.633Z"},"pt-BR":{"title":"Teste de 10 combinações de modelos/conexões na mesma tarefa Three.js","summary":"Um teste informal utilizou o mesmo prompt de construção de hangar futurista em Three.js para comparar o desempenho de geração de código de 10 combinações de modelos e conexões (harness). O GLM 5.3 Flash Max com OpenCode teve o menor tempo de execução (20 minutos) e a maior pontuação (96,89%), enquanto o Astra 6.0 Max com Codex levou mais tempo (37 minutos). O teste registrou o tempo de execução, o uso de tokens e a taxa de sucesso de chamadas de ferramentas para cada combinação.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Teste de 10 combinações de modelos/conexões na mesma tarefa Three.js - Aioga Notícias de IA","description":"Um teste informal utilizou o mesmo prompt de construção de hangar futurista em Three.js para comparar o desempenho de geração de código de 10 combinações de modelos e conexões (har...","url":"https://www.aioga.com/pt-BR/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:00:59.169Z"},"ru":{"title":"Я протестировал 10 комбинаций моделей/подключений на той же задаче Three.js","summary":"Неофициальное тестирование использовало один и тот же промпт для построения научно-фантастического ангара в Three.js, сравнивая производительность генерации кода 10 комбинаций моделей и подключений (harness). GLM 5.3 Flash Max вместе с OpenCode потратили наименьшее время (20 минут) и получили наивысший балл (96,89%), Astra 6.0 Max с Codex заняли больше всего времени (37 минут). В тестировании зарегистрированы время выполнения, расход токенов и успех вызова инструментов для каждой комбинации.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Я протестировал 10 комбинаций моделей/подключений на той же задаче Three.js - Aioga Новости ИИ","description":"Неофициальное тестирование использовало один и тот же промпт для построения научно-фантастического ангара в Three.js, сравнивая производительность генерации кода 10 комбинаций моде...","url":"https://www.aioga.com/ru/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:00.580Z"},"ar":{"title":"اختبرت 10 مجموعات من النماذج/الربط على نفس مهمة Three.js","summary":"اختبار غير رسمي استخدم نفس مقطع تحفيز بناء مستودع خيال علمي في Three.js، لمقارنة أداء توليد الكود بين 10 مجموعات من النماذج والربط (harness). كان GLM 5.3 Flash Max مع OpenCode الأسرع (20 دقيقة) والأعلى درجة (96.89%)، بينما استغرق Astra 6.0 Max مع Codex الأطول (37 دقيقة). سجل الاختبار الوقت المستغرق، وعدد التوكنات، ونسبة نجاح استدعاء الأدوات لكل مجموعة.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"اختبرت 10 مجموعات من النماذج/الربط على نفس مهمة Three.js - Aioga أخبار الذكاء الاصطناعي","description":"اختبار غير رسمي استخدم نفس مقطع تحفيز بناء مستودع خيال علمي في Three.js، لمقارنة أداء توليد الكود بين 10 مجموعات من النماذج والربط (harness). كان GLM 5.3 Flash Max مع OpenCode الأس...","url":"https://www.aioga.com/ar/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:07.504Z"},"hi":{"title":"मैंने एक ही Three.js कार्य पर 10 मॉडल/केबल संयोजनों का परीक्षण किया","summary":"एक अनौपचारिक परीक्षण 10 मॉडलों और हार्नेस संयोजनों में कोड जनरेशन प्रदर्शन की तुलना करने के लिए उसी Three.js विज्ञान-फाई गैरेज का उपयोग करके संकेत देता है। ओपनकोड के साथ जीएलएम 5.3 फ्लैश मैक्स ने सबसे कम समय (20 मिनट) लिया और उच्चतम (96.89%) स्कोर किया, जबकि कोडेक्स के साथ एस्ट्रा 6.0 मैक्स ने सबसे लंबा (37 मिनट) लिया। परीक्षण ने प्रत्येक संयोजन के लिए समय, टोकन उपयोग और टूल कॉल सफलता दर दर्ज की।","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"मैंने एक ही Three.js कार्य पर 10 मॉडल/केबल संयोजनों का परीक्षण किया - Aioga AI समाचार","description":"एक अनौपचारिक परीक्षण 10 मॉडलों और हार्नेस संयोजनों में कोड जनरेशन प्रदर्शन की तुलना करने के लिए उसी Three.js विज्ञान-फाई गैरेज का उपयोग करके संकेत देता है। ओपनकोड के साथ जीएलएम 5.3...","url":"https://www.aioga.com/hi/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:09.657Z"},"it":{"title":"Ho testato 10 combinazioni di modelli/connessioni sullo stesso compito Three.js","summary":"Un test informale ha utilizzato la stessa porzione di prompt per la costruzione di uno hangar fantascientifico in Three.js, confrontando le prestazioni di generazione di codice di 10 combinazioni di modelli e connessioni (harness). GLM 5.3 Flash Max con OpenCode ha impiegato meno tempo (20 minuti) e ottenuto il punteggio più alto (96,89%), mentre Astra 6.0 Max con Codex ha richiesto più tempo (37 minuti). Il test ha registrato il tempo impiegato, il numero di token usati e il tasso di successo delle chiamate agli strumenti per ogni combinazione.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Ho testato 10 combinazioni di modelli/connessioni sullo stesso compito Three.js - Aioga Notizie IA","description":"Un test informale ha utilizzato la stessa porzione di prompt per la costruzione di uno hangar fantascientifico in Three.js, confrontando le prestazioni di generazione di codice di...","url":"https://www.aioga.com/it/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:15.104Z"},"nl":{"title":"Ik testte 10 model/connection line combinaties op dezelfde Three.js taak","summary":"Een informele test gebruikte dezelfde Three.js sciencefiction hangarbouwwoordenset om de codegeneratieprestaties van 10 model- en connection line (harness) combinaties te vergelijken. GLM 5.3 Flash Max met OpenCode had de kortste tijd (20 minuten) en de hoogste score (96,89%), Astra 6.0 Max met Codex de langste tijd (37 minuten). De test registreerde tijdsduur, tokengebruik en succesratio van toolaanroepen voor elke combinatie.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Ik testte 10 model/connection line combinaties op dezelfde Three.js taak - Aioga AI-nieuws","description":"Een informele test gebruikte dezelfde Three.js sciencefiction hangarbouwwoordenset om de codegeneratieprestaties van 10 model- en connection line (harness) combinaties te vergelijk...","url":"https://www.aioga.com/nl/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:14.342Z"},"tr":{"title":"Aynı Three.js görevinde 10 model/bağlantı hattı kombinasyonunu test ettim","summary":"Resmi olmayan bir test, aynı Three.js bilim kurgu hangar yapılandırma promptunu kullanarak 10 model ile bağlantı hattı (harness) kombinasyonunun kod üretim performansını karşılaştırdı. GLM 5.3 Flash Max, OpenCode ile birlikte en kısa sürede (20 dakika) bitirdi ve en yüksek puanı (96,89%) aldı; Astra 6.0 Max Codex ile birlikte en uzun süreyi (37 dakika) aldı. Test, her kombinasyonun süre, token kullanımı ve araç çağrısı başarı oranını kaydetti.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Aynı Three.js görevinde 10 model/bağlantı hattı kombinasyonunu test ettim - Aioga AI Haberleri","description":"Resmi olmayan bir test, aynı Three.js bilim kurgu hangar yapılandırma promptunu kullanarak 10 model ile bağlantı hattı (harness) kombinasyonunun kod üretim performansını karşılaştı...","url":"https://www.aioga.com/tr/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:21.086Z"},"vi":{"title":"Tôi đã thử 10 kết hợp mô hình/dây nối trên cùng một nhiệm vụ Three.js","summary":"Một thử nghiệm không chính thức sử dụng cùng một đoạn gợi ý xây dựng kho cơ khí khoa học viễn tưởng Three.js, so sánh hiệu suất tạo mã của 10 kết hợp mô hình và dây nối (harness). GLM 5.3 Flash Max kết hợp OpenCode mất thời gian ngắn nhất (20 phút) và điểm cao nhất (96,89%), Astra 6.0 Max kết hợp Codex mất thời gian lâu nhất (37 phút). Thử nghiệm ghi lại thời gian, lượng token sử dụng và tỷ lệ thành công gọi công cụ của từng kết hợp.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Tôi đã thử 10 kết hợp mô hình/dây nối trên cùng một nhiệm vụ Three.js - Tin tức AI Aioga","description":"Một thử nghiệm không chính thức sử dụng cùng một đoạn gợi ý xây dựng kho cơ khí khoa học viễn tưởng Three.js, so sánh hiệu suất tạo mã của 10 kết hợp mô hình và dây nối (harness)....","url":"https://www.aioga.com/vi/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:22.967Z"},"id":{"title":"Saya menguji 10 kombinasi model/kabel pada tugas Three.js yang sama","summary":"Sebuah pengujian informal menggunakan prompt Build Sci-Fi Hangar Three.js yang sama, membandingkan performa generasi kode dari 10 kombinasi model dan kabel (harness). GLM 5.3 Flash Max dengan OpenCode membutuhkan waktu paling singkat (20 menit) dan memperoleh skor tertinggi (96,89%), Astra 6.0 Max dengan Codex memakan waktu paling lama (37 menit). Pengujian mencatat waktu yang dibutuhkan, jumlah token, dan tingkat keberhasilan panggilan alat dari setiap kombinasi.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Saya menguji 10 kombinasi model/kabel pada tugas Three.js yang sama - Berita AI Aioga","description":"Sebuah pengujian informal menggunakan prompt Build Sci-Fi Hangar Three.js yang sama, membandingkan performa generasi kode dari 10 kombinasi model dan kabel (harness). GLM 5.3 Flash...","url":"https://www.aioga.com/id/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:27.462Z"},"th":{"title":"ผมทดสอบการจับคู่โมเดล/สายเคเบิล 10 แบบในงาน Three.js เดียวกัน","summary":"การทดสอบแบบไม่เป็นทางการได้สร้างพรอมต์โดยใช้โรงรถไซไฟ Three.js เดียวกันเพื่อเปรียบเทียบประสิทธิภาพการสร้างโค้ดใน 10 รุ่นและการจับคู่ harnessGLM 5.3 Flash Max พร้อม OpenCode ใช้เวลาสั้นที่สุด (20 นาที) และได้คะแนนสูงสุด (96.89%) ขณะที่ Astra 6.0 Max กับ Codex ใช้เวลานานที่สุด (37 นาที)การทดสอบบันทึกเวลา การใช้โทเค็น และอัตราความสําเร็จในการเรียกใช้เครื่องมือสําหรับแต่ละชุด","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ผมทดสอบการจับคู่โมเดล/สายเคเบิล 10 แบบในงาน Three.js เดียวกัน - ข่าว AI Aioga","description":"การทดสอบแบบไม่เป็นทางการได้สร้างพรอมต์โดยใช้โรงรถไซไฟ Three.js เดียวกันเพื่อเปรียบเทียบประสิทธิภาพการสร้างโค้ดใน 10 รุ่นและการจับคู่ harnessGLM 5.3 Flash Max พร้อม OpenCode ใช้เวลา...","url":"https://www.aioga.com/th/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:32.033Z"},"pl":{"title":"Testowałem 10 kombinacji modeli/przewodów na tym samym zadaniu Three.js","summary":"Nieformalne testy wykorzystujące ten sam fragment promptu dotyczącego budowy science-fiction hangaru w Three.js porównywały wydajność generowania kodu 10 modeli z różnymi przewodami (harness). GLM 5.3 Flash Max w połączeniu z OpenCode zajęło najmniej czasu (20 minut) i uzyskało najwyższy wynik (96,89%), Astra 6.0 Max z Codex zajęło najwięcej czasu (37 minut). Test rejestrował czas trwania, wykorzystanie tokenów i skuteczność wywołań narzędzi dla każdej kombinacji.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Testowałem 10 kombinacji modeli/przewodów na tym samym zadaniu Three.js - Aioga Wiadomości AI","description":"Nieformalne testy wykorzystujące ten sam fragment promptu dotyczącego budowy science-fiction hangaru w Three.js porównywały wydajność generowania kodu 10 modeli z różnymi przewodam...","url":"https://www.aioga.com/pl/news/cmtspf1bu024xroou4hxsgbwc/","contentTranslated":true,"sourceHash":"07d82b836717112f","translatedAt":"2026-09-08T14:01:36.941Z"}},"evidenceTier":"verified-news","reviewStatus":"automated-ingest","indexable":true,"editorialCover":"/page-visuals/topic-timeline.png"}}