{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"Kimi K3 与 Fable 5 对比：开源模型在 93% 路由准确率下成本低至 50 倍","description":"在约 1，000 项智能体任务测试中，开源模型 Kimi K3 与闭源模型 Fable 5 整体准确率接近（SWE 基准 92.4% vs 92.6%），但 K3 在 Fireworks 上的推理成本可低至 Fable 的 1/50。通过任务路由，两者组合可实现 93% 的准确率，其中 K3 承担 72-96% 的任务流量，在保持前沿质量的同时大幅降低总成本。","url":"https://www.aioga.com/news/cmrvmzjnv03yobihbxui7w6qv/","mainEntityOfPage":"https://www.aioga.com/news/cmrvmzjnv03yobihbxui7w6qv/","datePublished":"2026-07-22T04:52:13.379Z","dateModified":"2026-07-22T04:52:13.379Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://fireworks.ai/blog/kimik3-fable","https://aihot.virxact.com/items/cmrvmzjnv03yobihbxui7w6qv"],"canonicalUrl":"https://www.aioga.com/news/cmrvmzjnv03yobihbxui7w6qv/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：在约 1，000 项智能体任务测试中，开源模型 Kimi K3 与闭源模型 Fable 5 整体准确率接近（SWE 基准 92.4% vs 92.6%），但 K3 在 Fireworks Aioga 将其归入「技巧观点」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrvmzjnv03yobihbxui7w6qv/","dateCreated":"2026-07-22T04:52:13.379Z","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":"fireworks.ai source article","url":"https://fireworks.ai/blog/kimik3-fable","datePublished":"2026-07-22T04:52:13.379Z","provider":{"@type":"Organization","name":"fireworks.ai","url":"https://fireworks.ai/blog/kimik3-fable"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrvmzjnv03yobihbxui7w6qv","datePublished":"2026-07-22T04:52:13.379Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrvmzjnv03yobihbxui7w6qv"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"fireworks.ai","url":"https://fireworks.ai/blog/kimik3-fable"},"article":{"id":"cmrvmzjnv03yobihbxui7w6qv","slug":"cmrvmzjnv03yobihbxui7w6qv","url":"https://www.aioga.com/news/cmrvmzjnv03yobihbxui7w6qv/","title":"Kimi K3 与 Fable 5 对比：开源模型在 93% 路由准确率下成本低至 50 倍","title_en":"Kimi K3 与《Fable》不相上下；Kimi K3 和《Fable》均属顶尖之作","summary":"在约 1，000 项智能体任务测试中，开源模型 Kimi K3 与闭源模型 Fable 5 整体准确率接近（SWE 基准 92.4% vs 92.6%），但 K3 在 Fireworks 上的推理成本可低至 Fable 的 1/50。通过任务路由，两者组合可实现 93% 的准确率，其中 K3 承担 72-96% 的任务流量，在保持前沿质量的同时大幅降低总成本。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://fireworks.ai/blog/kimik3-fable","aiHotUrl":"https://aihot.virxact.com/items/cmrvmzjnv03yobihbxui7w6qv","publishedAt":"2026-07-22T04:52:13.379Z","category":"技巧观点","score":57,"selected":false,"articleBody":["Announcing our Series D and $1B ARR","K3 is a frontier quality open model at a fraction of the cost. Even bigger is that it complements Fable predictably, which makes it possible to get the highest quality intelligence by routing tasks.","🧭 tl;dr: We ran Kimi K3 (open) against Fable 5 (closed) on ~1,000 agentic tasks finding:","We averaged benchmarks, each aimed at a different kind of work , and ran K3 and Fable 5 through the same harness. About 1,030 tasks in all, in real agent loops.","One quick definition before we get into the results. Oracle routing is a method for measuring the best theoretical performance by running the task through each model and then picking the cheapest correct option (the cost/performance ceiling). In a practical router, you don’t get to run your task against multiple models. The router makes a prediction of which model has the best cost and quality trade off, but ultimately it’s a guess.","In this study, oracle routing demonstrated K3 is selected for 72-96% of tasks. This suggests a near-perfect router might be achievable, by learning the difference between day-to-day tasks and the true long tail of frontier work. It will require an order of magnitude more routing data, and real world performance to say definitively.","From a 10,000 foot view, it can be easy to look at both models and call the head-to-head a tie. For example, if you look at SWE, the headline benchmark, K3 gets 92.4% , Fable 92.6% . Across the five types of tasks we benchmarked on, the two models tend to stay within a few points of each other, with Fable pulling slightly ahead on its coding-language breadth (Multi-lang).","It’s easy to stop there and say “they’re roughly even”. The news is that they have discretely better performance across different task types.","If you take a peek inside a single benchmark, there’s more to see than just a top-line accuracy number. Take SWE, where the two are dead even overall. If you split SWE by problem domain you can see where each model shines. K3 is sharpest on symbolic math and dev tooling; Fable wins on web & data visualization work. The same pattern runs through the multi-language set, where Fable's breadth carries Java, Python and C++, while K3 draws even on JavaScript and Rust.","For long-horizon work at a terminal, driving a shell and prodding at systems across dozens of turns, K3 showed its true colors. It cleared a batch of tasks Fable never cracked: a 7z hash, FEAL cryptanalysis, leaked secrets, a live vulnerability, runaway async jobs.","While quality is a near-tie at a high level, price isn't close.","So where's this huge price gap coming from? token pricing, prompt caching, and effort-per-task. On SWE for example, K3 works much harder than Fable: roughly 55 turns and 1.3M tokens a task versus 21 turns and 130K. On the long terminal tasks it's the other way around: Fable is the one that spirals, running up 64 turns and 1.5M tokens (sometimes straight into a timeout).","Prompt caching does most of the work of turning that effort into K3's price advantage: even when K3 reads ten times the tokens, with cache hits that means that SWE runs still come in lower cost than Fable. There’s a tradeoff. Tasks with extra turns generally mean more wall-clock time per run i.e. slower runs. If you need an answer in two seconds, that matters; if you're running agents in the background at scale, a bill that's a fraction of the size matters a lot more.","If you send every task to whoever handles it best, you don't land somewhere between the two models, you land above both.","Per-task routing always out performs any single model run:","The oracle router choose K3, 72-96% of task traffic. By architecting a router this way, you end up with overall quality above either model alone at a cost close to just using just the cost-optimized one.","Put both quality and cost on one plot. K3 in blue lands to the left (the more cost-effective side) of Fable in red in all five task-families. Accuracy trades back and forth: Fable pulls ahead on multi-language, K3 on terminal and legal, the rest roughly level.","Kimi K3 + Fable routed together unlocks their best qualities at the best price.","The single model provider, token maxxing days, are coming to an end. The task-level data says these models are specialists at very different prices. The best AI no longer comes out of a single lab, it’s a mixture of models."],"articleImages":[],"mediaStatus":"none","articleBodyZh":["宣布我们的D轮融资和10亿美元ARR","K3是一个前沿高质量的开源模型，成本仅为一小部分。更重要的是，它可以稳定地补充Fable，使得通过任务路由获得最高质量的智能成为可能。","🧭 简而言之：我们在约1,000个智能代理任务中，将Kimi K3（开源）与Fable 5（闭源）进行了对比，发现：","我们对每个基准进行了平均，每个基准针对不同类型的工作，并使用相同的框架运行K3和Fable 5。总共有约1,030个任务，在真实代理循环中进行。","在进入结果之前，先给出一个快速定义。Oracle路由是一种通过将任务在每个模型上运行后，选择最便宜的正确选项（成本/性能上限）来衡量最佳理论性能的方法。在实际路由中，你不能让任务同时运行在多个模型上。路由器会预测哪个模型在成本和质量上最优，但最终只是一个猜测。","在这项研究中，Oracle路由显示K3被选中的任务占72-96%。这表明，通过学习日常任务与前沿工作的真实长尾之间的差异，可能实现几乎完美的路由器。要得出明确结论，需要多一个数量级的路由数据和实际性能。","从10,000英尺的高度来看，很容易只看两个模型便认为正面对比是平手。例如，如果你看SWE这个主基准，K3得分92.4%，Fable得分92.6%。在我们基准测试的五种类型任务中，这两个模型的得分通常相差几个百分点，Fable在代码语言广度（多语言）上略占优势。","很容易就停止在这里，说“它们大致相当”。新闻是它们在不同任务类型上具有明显更好的性能。","如果你仔细查看单个基准，不只是看到一个顶线准确率就够了。以SWE为例，整体上两者完全相同。如果按问题领域拆分SWE，你可以看到每个模型的擅长领域。K3在符号数学和开发工具方面最为出色；Fable在网页和数据可视化工作上胜出。同样的模式在多语言集合中也存在，Fable在Java、Python和C++上具有广度优势，而K3在JavaScript和Rust上表现平分秋色。","对于在终端进行长周期工作的情况，如驱动 shell 并跨几十个回合操作系统，K3 展现了它的真实实力。它完成了一批 Fable 从未攻破的任务：7z 哈希、FEAL 密码分析、泄露的秘密、实时漏洞、失控的异步作业。","虽然在高水平上质量几乎不分伯仲，但价格却相差悬殊。","那么，这巨大的价格差距是从哪里来的呢？是令牌定价、提示缓存以及每个任务的处理工作量。例如在SWE上，K3 比 Fable 工作更努力：每个任务大约 55 回合和 130 万令牌，而 Fable 是 21 回合和 13 万令牌。在长终端任务上情况正好相反：Fable 才是耗费最多的，会达到 64 回合和 150 万令牌（有时直接导致超时）。","提示缓存大多负责将这些工作量转化为 K3 的价格优势：即使 K3 阅读的令牌数是 Fable 的十倍，但由于缓存命中，SWE 运行成本仍低于 Fable。这是一个权衡。增加回合数的任务通常意味着每次运行的实际时钟时间增加，即运行速度更慢。如果你需要在两秒内得到答案，这很重要；但如果你是在大规模后台运行代理，账单只是小部分，这就更重要了。","如果你将每个任务发送给最能处理它的模型，你并不会落在两种模型之间，而是超越两者。","按任务路由始终优于任何单一模型运行：","Oracle 路由器选择了 K3，占 72-96% 的任务流量。通过这样设计路由器，你最终获得的总体质量高于任何单一模型，成本接近仅使用最优化的模型。","将质量和成本放在同一图表上。K3 用蓝色表示，在所有五个任务类别中都位于 Fable（红色）左侧（更具成本效益的一侧）。准确率则此消彼长：Fable 在多语言上领先，K3 在终端和法律任务上领先，其他任务大致持平。","Kimi K3 + Fable 组合路由可以以最佳价格释放它们的最佳特性。","单一模型供应商、令牌最大化的时代即将结束。任务级数据表明，这些模型在不同价格下是专长不一的专家。最好的 AI 不再来自单一实验室，而是多模型的混合。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：在约 1，000 项智能体任务测试中，开源模型 Kimi K3 与闭源模型 Fable 5 整体准确率接近（SWE 基准 92.4% vs 92.6%），但 K3 在 Fireworks Aioga 将其归入「技巧观点」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T06:49:19.044Z","sourceHash":"6ab02431975260fb","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["技巧观点","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"Kimi K3 与 Fable 5 对比：开源模型在 93% 路由准确率下成本低至 50 倍","summary":"在约 1，000 项智能体任务测试中，开源模型 Kimi K3 与闭源模型 Fable 5 整体准确率接近（SWE 基准 92.4% vs 92.6%），但 K3 在 Fireworks 上的推理成本可低至 Fable 的 1/50。通过任务路由，两者组合可实现 93% 的准确率，其中 K3 承担 72-96% 的任务流量，在保持前沿质量的同时大幅降低总成本。","category":"技巧观点","source":"fireworks.ai","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3 与 Fable 5 对比：开源模型在 93% 路由准确率下成本低至 50 倍 - Aioga AI资讯","description":"在约 1，000 项智能体任务测试中，开源模型 Kimi K3 与闭源模型 Fable 5 整体准确率接近（SWE 基准 92.4% vs 92.6%），但 K3 在 Fireworks 上的推理成本可低至 Fable 的 1/50。通过任务路由，两者组合可实现 93% 的准确率，其中 K3 承担 72-96% 的任务流量，在保持前沿质量的同时大幅降低总成本...","url":"https://www.aioga.com/news/cmrvmzjnv03yobihbxui7w6qv/"},"en":{"title":"Comparison of Kimi K3 and Fable 5: Open-source model costs as low as 50 times under 93% routing accuracy","summary":"In approximately 1,000 agent task tests, the open-source model Kimi K3 and the closed-source model Fable 5 had similar overall accuracy (SWE benchmark 92.4% vs 92.6%), but K3's inference cost on Fireworks could be as low as 1/50 of Fable's. Through task routing, the combination of the two can achieve 93% accuracy, with K3 handling 72-96% of the task traffic, significantly reducing overall costs while maintaining cutting-edge quality.","category":"Insights","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Comparison of Kimi K3 and Fable 5: Open-source model costs as low as 50 times under 93% routing accuracy - Aioga AI News","description":"In approximately 1,000 agent task tests, the open-source model Kimi K3 and the closed-source model Fable 5 had similar overall accuracy (SWE benchmark 92.4% vs 92.6%), but K3's inf...","url":"https://www.aioga.com/en/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:22:05.446Z"},"ja":{"title":"Kimi K3 と Fable 5 の比較：オープンソースモデルは 93% のルーティング精度で費用を最大 50 倍削減","summary":"約1,000件のエージェントタスクテストにおいて、オープンソースモデルKimi K3とクローズドソースモデルFable 5の全体的な正確率はほぼ同じであった（SWEベンチマーク 92.4% vs 92.6%）。しかし、K3のFireworks上での推論コストはFableの1/50にまで低減できる。タスクルーティングを通じて、両者を組み合わせることで93%の正確率を達成でき、K3が72〜96%のタスクフローを担当することで、最先端の品質を維持しつつ総コストを大幅に削減できる。","category":"ヒントと視点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3 と Fable 5 の比較：オープンソースモデルは 93% のルーティング精度で費用を最大 50 倍削減 - Aioga AIニュース","description":"約1,000件のエージェントタスクテストにおいて、オープンソースモデルKimi K3とクローズドソースモデルFable 5の全体的な正確率はほぼ同じであった（SWEベンチマーク 92.4% vs 92.6%）。しかし、K3のFireworks上での推論コストはFableの1/50にまで低減できる。タスクルーティングを通じて、両者を組み合わせることで93%の正...","url":"https://www.aioga.com/ja/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:22:21.574Z"},"ko":{"title":"Kimi K3와 Fable 5 비교: 오픈 소스 모델이 93% 라우팅 정확도에서 비용을 최대 50배 절감","summary":"약 1,000개의 에이전트 과제 테스트에서, 오픈 소스 모델 Kimi K3와 폐쇄형 모델 Fable 5의 전체 정확도는 유사했습니다(SWE 벤치마크 92.4% vs 92.6%). 그러나 K3는 Fireworks에서의 추론 비용이 Fable의 1/50 수준으로 낮을 수 있습니다. 작업 라우팅을 통해 두 모델을 조합하면 93%의 정확도를 달성할 수 있으며, 이 과정에서 K3가 72-96%의 작업 흐름을 담당하여 최첨단 품질을 유지하면서 총 비용을 크게 줄일 수 있습니다.","category":"인사이트","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3와 Fable 5 비교: 오픈 소스 모델이 93% 라우팅 정확도에서 비용을 최대 50배 절감 - Aioga AI 뉴스","description":"약 1,000개의 에이전트 과제 테스트에서, 오픈 소스 모델 Kimi K3와 폐쇄형 모델 Fable 5의 전체 정확도는 유사했습니다(SWE 벤치마크 92.4% vs 92.6%). 그러나 K3는 Fireworks에서의 추론 비용이 Fable의 1/50 수준으로 낮을 수 있습니다. 작업 라우팅을 통해 두 모델을 조합하면 93...","url":"https://www.aioga.com/ko/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:23:10.464Z"},"es":{"title":"Comparación entre Kimi K3 y Fable 5: el modelo de código abierto tiene un costo hasta 50 veces menor con una precisión de enrutamiento del 93%","summary":"En aproximadamente 1,000 pruebas de tareas de agentes inteligentes, el modelo de código abierto Kimi K3 y el modelo cerrado Fable 5 tienen una precisión general similar (referencia SWE 92,4% vs 92,6%), pero el costo de inferencia de K3 en Fireworks puede ser tan bajo como 1/50 del de Fable. Mediante el enrutamiento de tareas, la combinación de ambos puede alcanzar una precisión del 93%, con K3 manejando entre el 72% y el 96% del flujo de tareas, reduciendo significativamente el costo total mientras se mantiene una calidad de vanguardia.","category":"Ideas","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Comparación entre Kimi K3 y Fable 5: el modelo de código abierto tiene un costo hasta 50 veces menor con una precisión de enrutamiento del 93% - Aioga Noticias de IA","description":"En aproximadamente 1,000 pruebas de tareas de agentes inteligentes, el modelo de código abierto Kimi K3 y el modelo cerrado Fable 5 tienen una precisión general similar (referencia...","url":"https://www.aioga.com/es/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:23:05.583Z"},"fr":{"title":"Comparaison entre Kimi K3 et Fable 5 : le modèle open source atteint un taux de précision du routage de 93 % pour un coût jusqu'à 50 fois inférieur","summary":"Dans environ 1 000 tests de tâches d'agents intelligents, le modèle open source Kimi K3 et le modèle propriétaire Fable 5 présentent une précision globale similaire (référence SWE 92,4 % contre 92,6 %), mais le coût de raisonnement de K3 sur Fireworks peut être aussi bas qu'1/50 de celui de Fable. Grâce au routage des tâches, la combinaison des deux peut atteindre une précision de 93 %, K3 prenant en charge 72 à 96 % du flux de tâches, ce qui permet de réduire considérablement le coût total tout en maintenant une qualité de pointe.","category":"Analyses","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Comparaison entre Kimi K3 et Fable 5 : le modèle open source atteint un taux de précision du routage de 93 % pour un coût jusqu'à 50 fois inférieur - Aioga Actualités IA","description":"Dans environ 1 000 tests de tâches d'agents intelligents, le modèle open source Kimi K3 et le modèle propriétaire Fable 5 présentent une précision globale similaire (référence SWE...","url":"https://www.aioga.com/fr/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:23:56.914Z"},"de":{"title":"Kimi K3 im Vergleich zu Fable 5: Open-Source-Modelle bei 93% Routengenauigkeit zu Kosten von bis zu 50-fach niedriger","summary":"In etwa 1.000 Agenten-Testaufgaben liegt die Gesamtgenauigkeit des Open-Source-Modells Kimi K3 nahe beim Closed-Source-Modell Fable 5 (SWE Benchmark 92,4 % vs. 92,6 %), aber die Inferenzkosten von K3 auf Fireworks können bis auf 1/50 von Fable reduziert werden. Durch Aufgaben-Routing kann die Kombination beider Modelle eine Genauigkeit von 93 % erreichen, wobei K3 72-96 % des Aufgabenaufkommens übernimmt und so die Gesamtkosten bei gleichbleibender Spitzenqualität erheblich senkt.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3 im Vergleich zu Fable 5: Open-Source-Modelle bei 93% Routengenauigkeit zu Kosten von bis zu 50-fach niedriger - Aioga KI-News","description":"In etwa 1.000 Agenten-Testaufgaben liegt die Gesamtgenauigkeit des Open-Source-Modells Kimi K3 nahe beim Closed-Source-Modell Fable 5 (SWE Benchmark 92,4 % vs. 92,6 %), aber die In...","url":"https://www.aioga.com/de/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:23:57.617Z"},"pt-BR":{"title":"Comparação entre Kimi K3 e Fable 5: modelo de código aberto com custo até 50 vezes menor a 93% de precisão de roteamento","summary":"Em cerca de 1.000 testes de tarefas de agentes inteligentes, o modelo open-source Kimi K3 e o modelo closed-source Fable 5 têm taxas de precisão gerais semelhantes (benchmark SWE 92,4% vs 92,6%), mas o custo de inferência do K3 no Fireworks pode ser até 1/50 do Fable. Através do roteamento de tarefas, a combinação dos dois pode atingir 93% de precisão, com o K3 assumindo 72-96% do fluxo de tarefas, reduzindo significativamente o custo total enquanto mantém qualidade de ponta.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Comparação entre Kimi K3 e Fable 5: modelo de código aberto com custo até 50 vezes menor a 93% de precisão de roteamento - Aioga Notícias de IA","description":"Em cerca de 1.000 testes de tarefas de agentes inteligentes, o modelo open-source Kimi K3 e o modelo closed-source Fable 5 têm taxas de precisão gerais semelhantes (benchmark SWE 9...","url":"https://www.aioga.com/pt-BR/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:24:39.795Z"},"ru":{"title":"Сравнение Kimi K3 и Fable 5: открытая модель при точности маршрутизации 93% может снижать затраты до 50 раз","summary":"В тестировании примерно 1 000 задач агентов открытая модель Kimi K3 и закрытая модель Fable 5 имеют схожую общую точность (бенчмарк SWE 92,4% против 92,6%), но стоимость рассуждений K3 на Fireworks может быть в 50 раз ниже, чем у Fable. С помощью маршрутизации задач их комбинация может достигать точности 93%, при этом K3 выполняет 72–96% потоков задач, значительно снижая общие расходы при сохранении передового качества.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Сравнение Kimi K3 и Fable 5: открытая модель при точности маршрутизации 93% может снижать затраты до 50 раз - Aioga Новости ИИ","description":"В тестировании примерно 1 000 задач агентов открытая модель Kimi K3 и закрытая модель Fable 5 имеют схожую общую точность (бенчмарк SWE 92,4% против 92,6%), но стоимость рассуждени...","url":"https://www.aioga.com/ru/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:24:45.674Z"},"ar":{"title":"مقارنة بين Kimi K3 و Fable 5: نموذج مفتوح المصدر بتكلفة منخفضة تصل إلى 50 مرة مع دقة مسار 93%","summary":"في حوالي 1,000 اختبار لمهام الوكلاء الذكيين، كان دقة النموذج المفتوح المصدر Kimi K3 والنموذج المغلق المصدر Fable 5 متقاربة بشكل عام (المعيار SWE 92.4٪ مقابل 92.6٪)، لكن تكلفة الاستدلال لـ K3 على Fireworks يمكن أن تصل إلى 1/50 من تكلفة Fable. من خلال توجيه المهام، يمكن لمزيجهما تحقيق دقة بنسبة 93٪، حيث يتولى K3 ما بين 72-96٪ من حركة مرور المهام، مما يقلل بشكل كبير من التكلفة الإجمالية مع الحفاظ على الجودة المتقدمة.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"مقارنة بين Kimi K3 و Fable 5: نموذج مفتوح المصدر بتكلفة منخفضة تصل إلى 50 مرة مع دقة مسار 93% - Aioga أخبار الذكاء الاصطناعي","description":"في حوالي 1,000 اختبار لمهام الوكلاء الذكيين، كان دقة النموذج المفتوح المصدر Kimi K3 والنموذج المغلق المصدر Fable 5 متقاربة بشكل عام (المعيار SWE 92.4٪ مقابل 92.6٪)، لكن تكلفة الاست...","url":"https://www.aioga.com/ar/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:25:30.417Z"},"hi":{"title":"Kimi K3 और Fable 5 की तुलना: ओपन-सोर्स मॉडल 93% रूटिंग सटीकता पर लागत को 50 गुना तक घटाता है","summary":"लगभग 1,000 एजेंट कार्य परीक्षणों में, ओपन-सोर्स मॉडल Kimi K3 और क्लोज़्ड-सोर्स मॉडल Fable 5 की कुल सटीकता करीबी थी (SWE बेंचमार्क 92.4% बनाम 92.6%), लेकिन K3 की Fireworks पर अनुमान लागत Fable की 1/50 तक कम हो सकती है। कार्य रूटिंग के माध्यम से, दोनों का संयोजन 93% की सटीकता प्राप्त कर सकता है, जिसमें K3 72-96% कार्य यातायात संभालता है, जिससे अग्रणी गुणवत्ता बनाए रखते हुए कुल लागत में काफी कमी आती है।","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3 और Fable 5 की तुलना: ओपन-सोर्स मॉडल 93% रूटिंग सटीकता पर लागत को 50 गुना तक घटाता है - Aioga AI समाचार","description":"लगभग 1,000 एजेंट कार्य परीक्षणों में, ओपन-सोर्स मॉडल Kimi K3 और क्लोज़्ड-सोर्स मॉडल Fable 5 की कुल सटीकता करीबी थी (SWE बेंचमार्क 92.4% बनाम 92.6%), लेकिन K3 की Fireworks पर अनुमान...","url":"https://www.aioga.com/hi/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:25:36.639Z"},"it":{"title":"Confronto tra Kimi K3 e Fable 5: il modello open source ha un costo fino a 50 volte inferiore con una precisione di routing del 93%","summary":"In circa 1.000 test di compiti per agenti intelligenti, il modello open source Kimi K3 e il modello closed source Fable 5 hanno mostrato un tasso di accuratezza complessivo simile (benchmark SWE 92,4% vs 92,6%), ma il costo di inferenza di K3 su Fireworks può essere fino a 1/50 di quello di Fable. Attraverso l'instradamento dei compiti, la combinazione dei due può raggiungere un'accuratezza del 93%, con K3 che gestisce il 72-96% del flusso di compiti, riducendo significativamente il costo totale pur mantenendo una qualità all'avanguardia.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Confronto tra Kimi K3 e Fable 5: il modello open source ha un costo fino a 50 volte inferiore con una precisione di routing del 93% - Aioga Notizie IA","description":"In circa 1.000 test di compiti per agenti intelligenti, il modello open source Kimi K3 e il modello closed source Fable 5 hanno mostrato un tasso di accuratezza complessivo simile...","url":"https://www.aioga.com/it/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:26:18.876Z"},"nl":{"title":"Kimi K3 versus Fable 5: open source model met 93% routeringsnauwkeurigheid en kosten tot 50 keer lager","summary":"In ongeveer 1.000 agent-taken tests, waren het open-source model Kimi K3 en het closed-source model Fable 5 qua algehele nauwkeurigheid vergelijkbaar (SWE benchmark 92,4% versus 92,6%), maar de inferentiekosten van K3 op Fireworks kunnen oplopen tot slechts 1/50 van die van Fable. Door taakroutering te gebruiken, kan een combinatie van beide een nauwkeurigheid van 93% bereiken, waarbij K3 72-96% van het taakverkeer afhandelt, waardoor de totale kosten aanzienlijk worden verlaagd terwijl de toonaangevende kwaliteit behouden blijft.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3 versus Fable 5: open source model met 93% routeringsnauwkeurigheid en kosten tot 50 keer lager - Aioga AI-nieuws","description":"In ongeveer 1.000 agent-taken tests, waren het open-source model Kimi K3 en het closed-source model Fable 5 qua algehele nauwkeurigheid vergelijkbaar (SWE benchmark 92,4% versus 92...","url":"https://www.aioga.com/nl/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:26:17.972Z"},"tr":{"title":"Kimi K3 ve Fable 5 Karşılaştırması: Açık kaynak modeli %93 yönlendirme doğruluğunda maliyeti 50 kata kadar düşük","summary":"Yaklaşık 1.000 ajan görevi testinde, açık kaynak modeli Kimi K3 ile kapalı kaynak modeli Fable 5'in genel doğruluk oranları yakın (SWE kriteri 92,4% vs 92,6%), ancak K3'ün Fireworks üzerindeki çıkarım maliyeti Fable'ın 1/50'si kadar düşük olabiliyor. Görev yönlendirmesi sayesinde, ikisinin birleşimi %93 doğruluk oranına ulaşabilir; K3, görev trafiğinin %72-96'sını üstlenerek, öncü kaliteyi korurken toplam maliyeti önemli ölçüde azaltır.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kimi K3 ve Fable 5 Karşılaştırması: Açık kaynak modeli %93 yönlendirme doğruluğunda maliyeti 50 kata kadar düşük - Aioga AI Haberleri","description":"Yaklaşık 1.000 ajan görevi testinde, açık kaynak modeli Kimi K3 ile kapalı kaynak modeli Fable 5'in genel doğruluk oranları yakın (SWE kriteri 92,4% vs 92,6%), ancak K3'ün Firework...","url":"https://www.aioga.com/tr/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:27:12.515Z"},"vi":{"title":"So sánh Kimi K3 và Fable 5: Mô hình mã nguồn mở có chi phí thấp hơn tới 50 lần với độ chính xác định tuyến 93%","summary":"Trong khoảng 1.000 bài kiểm tra tác vụ của đại lý thông minh, mô hình mã nguồn mở Kimi K3 và mô hình đóng nguồn Fable 5 có độ chính xác tổng thể gần như nhau (chuẩn SWE 92,4% so với 92,6%), nhưng chi phí suy luận của K3 trên Fireworks có thể thấp đến 1/50 so với Fable. Thông qua định tuyến tác vụ, sự kết hợp của cả hai mô hình có thể đạt độ chính xác 93%, trong đó K3 đảm nhận 72-96% lưu lượng tác vụ, đồng thời giảm đáng kể tổng chi phí trong khi vẫn duy trì chất lượng tiên tiến.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"So sánh Kimi K3 và Fable 5: Mô hình mã nguồn mở có chi phí thấp hơn tới 50 lần với độ chính xác định tuyến 93% - Tin tức AI Aioga","description":"Trong khoảng 1.000 bài kiểm tra tác vụ của đại lý thông minh, mô hình mã nguồn mở Kimi K3 và mô hình đóng nguồn Fable 5 có độ chính xác tổng thể gần như nhau (chuẩn SWE 92,4% so vớ...","url":"https://www.aioga.com/vi/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:27:02.920Z"},"id":{"title":"Perbandingan Kimi K3 dengan Fable 5: Model sumber terbuka memiliki biaya hingga 50 kali lebih rendah dengan akurasi rute 93%","summary":"Dalam uji coba sekitar 1.000 tugas agen cerdas, model open-source Kimi K3 memiliki tingkat akurasi keseluruhan yang hampir sama dengan model closed-source Fable 5 (benchmark SWE 92,4% vs 92,6%), namun biaya inferensi K3 di Fireworks bisa serendah 1/50 dari Fable. Melalui perutean tugas, kombinasi keduanya dapat mencapai akurasi 93%, di mana K3 menangani 72-96% dari aliran tugas, secara signifikan mengurangi biaya total sambil mempertahankan kualitas terdepan.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Perbandingan Kimi K3 dengan Fable 5: Model sumber terbuka memiliki biaya hingga 50 kali lebih rendah dengan akurasi rute 93% - Berita AI Aioga","description":"Dalam uji coba sekitar 1.000 tugas agen cerdas, model open-source Kimi K3 memiliki tingkat akurasi keseluruhan yang hampir sama dengan model closed-source Fable 5 (benchmark SWE 92...","url":"https://www.aioga.com/id/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:27:51.767Z"},"th":{"title":"เปรียบเทียบ Kimi K3 กับ Fable 5: โมเดลโอเพนซอร์สมีค่าใช้จ่ายต่ำถึง 50 เท่าที่ความแม่นยำการกำหนดเส้นทาง 93%","summary":"ในการทดสอบงานเอเย่นต์อัจฉริยะประมาณ 1,000 งาน โมเดลโอเพ่นซอร์ส Kimi K3 และโมเดลปิด Fable 5 มีความแม่นยำโดยรวมใกล้เคียงกัน (มาตรฐาน SWE 92.4% เทียบกับ 92.6%) แต่ K3 มีต้นทุนการคำนวณใน Fireworks ต่ำกว่าของ Fable ถึง 1/50 ผ่านการกำหนดเส้นทางงาน การรวมทั้งสองสามารถทำให้ได้ความแม่นยำ 93% โดยที่ K3 รับผิดชอบงาน 72-96% ของการไหลงาน ในขณะที่ยังคงคุณภาพระดับแนวหน้าและลดต้นทุนรวมอย่างมาก","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"เปรียบเทียบ Kimi K3 กับ Fable 5: โมเดลโอเพนซอร์สมีค่าใช้จ่ายต่ำถึง 50 เท่าที่ความแม่นยำการกำหนดเส้นทาง 93% - ข่าว AI Aioga","description":"ในการทดสอบงานเอเย่นต์อัจฉริยะประมาณ 1,000 งาน โมเดลโอเพ่นซอร์ส Kimi K3 และโมเดลปิด Fable 5 มีความแม่นยำโดยรวมใกล้เคียงกัน (มาตรฐาน SWE 92.4% เทียบกับ 92.6%) แต่ K3 มีต้นทุนการคำนวณ...","url":"https://www.aioga.com/th/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:28:01.494Z"},"pl":{"title":"Porównanie Kimi K3 i Fable 5: model open-source przy 93% dokładności routingu kosztuje nawet 50 razy mniej","summary":"W testach około 1 000 zadań agentów, otwarty model Kimi K3 i zamknięty model Fable 5 osiągnęły zbliżoną ogólną dokładność (benchmark SWE 92,4% vs 92,6%), jednak koszt wnioskowania K3 na Fireworks może być aż 50 razy niższy niż Fable. Dzięki trasowaniu zadań, połączenie obu modeli może osiągnąć 93% dokładności, przy czym K3 obsługuje 72-96% ruchu zadań, znacznie obniżając całkowity koszt przy zachowaniu najwyższej jakości.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Porównanie Kimi K3 i Fable 5: model open-source przy 93% dokładności routingu kosztuje nawet 50 razy mniej - Aioga Wiadomości AI","description":"W testach około 1 000 zadań agentów, otwarty model Kimi K3 i zamknięty model Fable 5 osiągnęły zbliżoną ogólną dokładność (benchmark SWE 92,4% vs 92,6%), jednak koszt wnioskowania...","url":"https://www.aioga.com/pl/news/cmrvmzjnv03yobihbxui7w6qv/","contentTranslated":true,"sourceHash":"94f4b1de197f9794","translatedAt":"2026-07-23T00:28:51.949Z"}}}}