{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-09-15T00:02:40.869Z","headline":"Real-SWE 发布：基于私有企业代码库评测前沿编码模型，最高解决率 38.8%","description":"Specific 发布 Real-SWE 基准，用从真实公司授权的私有生产代码库任务评测前沿编码模型。","url":"https://www.aioga.com/news/cmtz5c0hd0ujmroup4uua5duy/","mainEntityOfPage":"https://www.aioga.com/news/cmtz5c0hd0ujmroup4uua5duy/","datePublished":"2026-09-13T01:18:53.000Z","dateModified":"2026-09-13T01:18:53.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://withspecific.com/benchmarks/real-swe","https://aihot.news/items/cmtz5c0hd0ujmroup4uua5duy"],"canonicalUrl":"https://www.aioga.com/news/cmtz5c0hd0ujmroup4uua5duy/","directAnswer":{"@type":"Answer","text":"Specific 发布 Real-SWE 基准，使用经真实公司授权的私有生产代码库任务评测前沿 AI 模型。标题显示，当前最高解决率为 38.8%。","url":"https://www.aioga.com/news/cmtz5c0hd0ujmroup4uua5duy/","dateCreated":"2026-09-13T01:18:53.000Z","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":"withspecific.com source article","url":"https://withspecific.com/benchmarks/real-swe","datePublished":"2026-09-13T01:18:53.000Z","provider":{"@type":"Organization","name":"withspecific.com","url":"https://withspecific.com/benchmarks/real-swe"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/cmtz5c0hd0ujmroup4uua5duy","datePublished":"2026-09-13T01:18:53.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/cmtz5c0hd0ujmroup4uua5duy"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"withspecific.com","url":"https://withspecific.com/benchmarks/real-swe"},"geoDeepAnswer":null,"article":{"id":"cmtz5c0hd0ujmroup4uua5duy","slug":"cmtz5c0hd0ujmroup4uua5duy","url":"https://www.aioga.com/news/cmtz5c0hd0ujmroup4uua5duy/","title":"Real-SWE 发布：基于私有企业代码库评测前沿编码模型，最高解决率 38.8%","title_en":"","summary":"Specific 发布 Real-SWE 基准，用从真实公司授权的私有生产代码库任务评测前沿编码模型。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://withspecific.com/benchmarks/real-swe","aiHotUrl":"https://aihot.news/items/cmtz5c0hd0ujmroup4uua5duy","publishedAt":"2026-09-13T01:18:53.000Z","category":"行业动态","score":58,"selected":false,"articleBody":["Benchmarking frontier AI models on private, real-world, enterprise codebases.","Today we are releasing Real-SWE, a benchmark that evaluates frontier AI models on private, real-world, enterprise codebases. Each task comes from a private production codebase that we licensed from a real-world company. These are problems their engineers work on, with all the context and complexity that comes with an existing product.","Can a coding agent actually do the work of a software engineer in the real world?","Expert-generated or synthetic tasks can be well designed, but they aren’t the verbatim, actual tasks that engineers in real companies need to do. Our tasks differ on two axes: the underlying coding artifact and specificity of the instruction. Both add complexities that challenge today’s frontier models.","We use native harnesses to reflect how enterprise engineers work in practice, evaluating model-and-harness combinations rather than models in isolation.","Fix invoice billing so each business charges the right tax and exempt customers aren't taxed.","Billing reopens on Monday and every invoice this service issues is coming out untaxed. Each business on the platform settles its tax a different way: some maintain a rate themselves, some want each invoice priced against the buyer's destination by our tax authority provider, and some collect nothing at all, while a customer we hold an exemption for is charged nothing whichever way its business is configured. Pricing a destination means going to the authority with both addresses, the priced lines and the product category that business sells under, on the sandbox or the production authority according to the account the business is on; an address the authority refuses must be reported without stopping the invoice. The rate, the tax and the gross belong on the issued invoice, and once an invoice is settled the sale is filed back to the authority under that invoice's number so the returns reconcile. Invoices between European parties show both sides' VAT registrations. The authority and ledger are available at TAX_JAR_URL , PROD_TAX_JAR_URL and INFLUX_URL .","Tools and services across Real-SWE task environments. Each task exposes only the services its workflow needs.","We selected codebases through a rigorous screening process, focusing on real companies with substantial usage, strong engineering teams, and demanding production workloads. The sample tasks analyzed below come from these codebases, including:","We prioritize code written to meet an actual user or business need over code written solely to create a benchmark task. Production engineering requires understanding existing architecture, preserving behavior that users rely on, and making changes within real operational constraints.","Our tasks describe the change needed, leaving agents to discover implementation details in the codebase and surrounding tools. Any behavior required by the verifier must be stated or reasonably discoverable. This leads to our prompts being slightly underspecified, about par with DeepSWE and Terminal Bench, but specific enough to not omit instructions.","The work is cross-functional and complex: a single change can span multiple parts of the application. Agents must understand existing business logic and company coding patterns while keeping the surrounding system working.","A typical Real-SWE instruction is 1,742 characters.","11 files in Real-SWE, compared with 6 in FrontierCode and DeepSWE.","71.4 % of rollouts under 10 minutes failed, compared with 73.4 % of longer rollouts.","Triaging multiple systems and understanding requirements in codebases riddled with existing business logic and coding patterns is difficult.","Every task is inspired or lifted verbatim from a private, real-world codebase. We find these types of tasks super interesting for three reasons:","Here's an analysis of a small sample of tasks from our benchmark. If you're interested in the sample, request access here：/benchmarks/real-swe/request-access.","Select a task to view model results. Percentages show the overall resolution rate.","Failures are grouped by observed submission behavior using the same taxonomy across models, following DeepSWE：https://arxiv.org/html/2607.07946v1#A3.","One square per rollout: each row is a task, each column a trial, eight trials per task for every model.","Percentages are out of each model's failed runs, not all runs.","Builds on a guess about the system instead of checking it in the workspace.","Leaves out behavior the instruction requires.","Right idea, wired into the surrounding system incorrectly.","Breaks existing behavior while making the change.","Delivers the change somewhere the running application never calls, such as a one-off script.","Each agent was run in an isolated sandbox. All tasks are in Harbor format, and verifiers are injected at grading time. The verifiers are inspired by existing test suites in the codebase or use those tests verbatim."],"articleImages":[{"sourceUrl":"https://withspecific.com/logos/zai.svg?dpl=dpl_ABGtodWDkmm7BqTikqe2cpVTakLw","alt":"","afterParagraph":2,"url":"/media/articles/cmtz114lu0pforoupt1npkc1i/c337570766e65cde.jpg"}],"mediaStatus":"ok","articleBodyZh":["在私有、真实世界的企业代码库上对前沿 AI 模型进行基准测试。","今天我们发布了 Real-SWE，这是一个在私有、真实世界的企业代码库上评估前沿 AI 模型的基准。每个任务都来自我们从真实公司许可的私有生产代码库。这些是他们工程师正在处理的问题，具有现有产品所带来的所有上下文和复杂性。","在现实世界中，编码代理真的能替代软件工程师的工作吗？","专家生成的或合成的任务可以设计得很好，但它们并不是工程师在真实公司中需要完成的逐字实际任务。我们的任务在两个方面有所不同：基础代码工件和指令的具体性。这两者都增加了挑战当今前沿模型的复杂性。","我们使用本地测试环境来反映企业工程师的实际工作方式，评估模型和测试环境的组合，而不是单独评估模型。","修复发票计费，使每个业务收取正确的税款，并确保免税客户不被征税。","计费将在周一重新开启，本服务发出的每张发票都会没有税款。平台上的每个业务以不同方式结算税款：有些自行维护税率，有些希望每张发票根据买方所在地由我们的税务提供者定价，还有些不收取任何税款，而对于我们持有免税资格的客户，无论业务配置如何，都不收取费用。为目的地定价意味着根据业务所在账户，向税务机关提供双方地址、定价项目以及业务销售的产品类别，在沙箱或生产税务环境中进行；税务机关拒绝的地址必须报告，但不能阻止发票开具。税率、税金和总额必须出现在已开具的发票上，一旦发票结算，销售额会根据该发票号回报给税务机关，以便汇总申报。欧洲双方之间的发票显示双方的增值税登记号码。税务机关和账簿可在 TAX_JAR_URL、PROD_TAX_JAR_URL 和 INFLUX_URL 查阅。","Real-SWE 任务环境中的工具和服务。每个任务仅公开其工作流所需的服务。","我们通过严格的筛选过程选择了代码库，重点关注那些拥有大量使用、强大工程团队和高强度生产工作负载的真实公司。下面分析的样本任务来自这些代码库，包括：","我们优先考虑为满足实际用户或业务需求而编写的代码，而不是仅为了创建基准任务而编写的代码。生产工程需要理解现有架构，保留用户依赖的行为，并在真实操作约束下进行更改。","我们的任务描述所需的更改，留给代理去发现代码库及相关工具中的实现细节。验证者要求的任何行为必须明确说明或可以合理地发现。这使我们的提示略显不完全明确，与 DeepSWE 和 Terminal Bench 相当，但足够具体，不会遗漏指令。","这项工作是跨职能且复杂的：单一更改可能涉及应用程序的多个部分。代理必须理解现有业务逻辑和公司编码模式，同时保持周围系统的正常运行。","一次典型的真实SWE指令是1,742个字符。","Real-SWE 中有 11 个文件，相比之下 FrontierCode 和 DeepSWE 有 6 个文件。","71.4% 的 10 分钟以内的部署失败，而较长时间部署失败率为 73.4%。","在充满现有业务逻辑和编码模式的代码库中，对多个系统进行分类和理解需求是困难的。","每个任务都源自或直接摘自私有的真实世界代码库。我们发现这类任务非常有趣，原因有三：","以下是我们基准中的一小部分任务分析。如果您对样本感兴趣，请在这里申请访问：/benchmarks/real-swe/request-access。","选择任务以查看模型结果。百分比显示整体解决率。","失败按提交行为进行分组，使用跨模型的相同分类法，遵循 DeepSWE：https://arxiv.org/html/2607.07946v1#A3。","每次部署一个方格：每行代表一个任务，每列代表一次试验，每个任务的每个模型有八次试验。","百分比基于每个模型的失败运行，而不是所有运行。","建立在对系统的猜测上，而不是在工作区中进行检查。","遗漏指令要求的行为。","思路正确，但与周围系统的连接不正确。","在进行更改时破坏了现有行为。","将更改交付到运行的应用程序从未调用的地方，例如一次性脚本。","每个代理都在隔离的沙箱中运行。所有任务都是 Harbor 格式，验证器在评分时注入。验证器的灵感来自代码库中现有的测试套件，或者按原样使用这些测试。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Specific 发布 Real-SWE 基准，使用经真实公司授权的私有生产代码库任务评测前沿 AI 模型。标题显示，当前最高解决率为 38.8%。","background":"该基准强调任务来自真实企业生产代码库，并保留现有产品中的上下文与复杂性。评测采用原生 harness，考察模型与 harness 的组合，而非只单独比较模型。","viewpoint":"Aioga 判断：Real-SWE 的核心价值在于把评测对象从设计或合成任务，推进到来源于实际生产代码库的具体工程问题；但其结果仍应结合任务与环境理解。","implications":"可能影响：企业代码任务的评测结果可能更能反映特定工作流中的表现，但单一基准不足以代表模型能够完成所有软件工程工作；比较时需要关注任务、代码库和 harness 条件。","nextStep":"后续观察：应关注 Real-SWE 后续披露的任务覆盖范围、评分细节及不同模型与 harness 组合的结果，以判断 38.8% 解决率的适用边界。","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":"2026-09-13T02:29:46.399Z","sourceHash":"5a4e9c8ae9d32e56","review":{"approved":true,"groundedness":96,"clarity":92,"duplicationRisk":8,"blockingIssues":[],"notes":["“当前最高解决率为 38.8%”有标题来源支持，但来源摘录未提供该数值对应的模型、任务范围或评分细节；后续内容已适当提示其适用边界。","“核心价值”及“可能更能反映”等属于明确标注的判断或谨慎推论，不构成观点冒充事实。"]},"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":"Real-SWE 发布：在企业私有真实代码库上评测前沿编程模型","summary":"Specific 团队发布 Real-SWE 基准，任务取自经授权的真实公司私有生产代码库，评测 8 个前沿模型与 harness 组合的解决率。","category":"行业动态","source":"withspecific.com","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE 发布：在企业私有真实代码库上评测前沿编程模型 - Aioga AI资讯","description":"Specific 团队发布 Real-SWE 基准，任务取自经授权的真实公司私有生产代码库，评测 8 个前沿模型与 harness 组合的解决率。","url":"https://www.aioga.com/news/cmtz5c0hd0ujmroup4uua5duy/","articleBody":["在私有、真实世界的企业代码库上对前沿 AI 模型进行基准测试。","今天我们发布了 Real-SWE，这是一个在私有、真实世界的企业代码库上评估前沿 AI 模型的基准。每个任务都来自我们从真实公司许可的私有生产代码库。这些是他们工程师正在处理的问题，具有现有产品所带来的所有上下文和复杂性。","在现实世界中，编码代理真的能替代软件工程师的工作吗？","专家生成的或合成的任务可以设计得很好，但它们并不是工程师在真实公司中需要完成的逐字实际任务。我们的任务在两个方面有所不同：基础代码工件和指令的具体性。这两者都增加了挑战当今前沿模型的复杂性。","我们使用本地测试环境来反映企业工程师的实际工作方式，评估模型和测试环境的组合，而不是单独评估模型。","修复发票计费，使每个业务收取正确的税款，并确保免税客户不被征税。","计费将在周一重新开启，本服务发出的每张发票都会没有税款。平台上的每个业务以不同方式结算税款：有些自行维护税率，有些希望每张发票根据买方所在地由我们的税务提供者定价，还有些不收取任何税款，而对于我们持有免税资格的客户，无论业务配置如何，都不收取费用。为目的地定价意味着根据业务所在账户，向税务机关提供双方地址、定价项目以及业务销售的产品类别，在沙箱或生产税务环境中进行；税务机关拒绝的地址必须报告，但不能阻止发票开具。税率、税金和总额必须出现在已开具的发票上，一旦发票结算，销售额会根据该发票号回报给税务机关，以便汇总申报。欧洲双方之间的发票显示双方的增值税登记号码。税务机关和账簿可在 TAX_JAR_URL、PROD_TAX_JAR_URL 和 INFLUX_URL 查阅。","Real-SWE 任务环境中的工具和服务。每个任务仅公开其工作流所需的服务。","我们通过严格的筛选过程选择了代码库，重点关注那些拥有大量使用、强大工程团队和高强度生产工作负载的真实公司。下面分析的样本任务来自这些代码库，包括：","我们优先考虑为满足实际用户或业务需求而编写的代码，而不是仅为了创建基准任务而编写的代码。生产工程需要理解现有架构，保留用户依赖的行为，并在真实操作约束下进行更改。","我们的任务描述所需的更改，留给代理去发现代码库及相关工具中的实现细节。验证者要求的任何行为必须明确说明或可以合理地发现。这使我们的提示略显不完全明确，与 DeepSWE 和 Terminal Bench 相当，但足够具体，不会遗漏指令。","这项工作是跨职能且复杂的：单一更改可能涉及应用程序的多个部分。代理必须理解现有业务逻辑和公司编码模式，同时保持周围系统的正常运行。","一次典型的真实SWE指令是1,742个字符。","Real-SWE 中有 11 个文件，相比之下 FrontierCode 和 DeepSWE 有 6 个文件。","71.4% 的 10 分钟以内的部署失败，而较长时间部署失败率为 73.4%。","在充满现有业务逻辑和编码模式的代码库中，对多个系统进行分类和理解需求是困难的。","每个任务都源自或直接摘自私有的真实世界代码库。我们发现这类任务非常有趣，原因有三：","以下是我们基准中的一小部分任务分析。如果您对样本感兴趣，请在这里申请访问：/benchmarks/real-swe/request-access。","选择任务以查看模型结果。百分比显示整体解决率。","失败按提交行为进行分组，使用跨模型的相同分类法，遵循 DeepSWE：https://arxiv.org/html/2607.07946v1#A3。","每次部署一个方格：每行代表一个任务，每列代表一次试验，每个任务的每个模型有八次试验。","百分比基于每个模型的失败运行，而不是所有运行。","建立在对系统的猜测上，而不是在工作区中进行检查。","遗漏指令要求的行为。","思路正确，但与周围系统的连接不正确。","在进行更改时破坏了现有行为。","将更改交付到运行的应用程序从未调用的地方，例如一次性脚本。","每个代理都在隔离的沙箱中运行。所有任务都是 Harbor 格式，验证器在评分时注入。验证器的灵感来自代码库中现有的测试套件，或者按原样使用这些测试。"]},"en":{"title":"Real-SWE Released: Evaluating Frontier Coding Models on Private Enterprise Codebases, Highest Solution Rate 38.8%","summary":"Specific has released the Real-SWE benchmark, which evaluates frontier coding models using tasks from real private production codebases licensed from companies.","category":"Industry","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE Released: Evaluating Frontier Coding Models on Private Enterprise Codebases, Highest Solution Rate 38.8% - Aioga AI News","description":"Specific has released the Real-SWE benchmark, which evaluates frontier coding models using tasks from real private production codebases licensed from companies.","url":"https://www.aioga.com/en/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:15.744Z"},"ja":{"title":"Real-SWEリリース:民間企業のコードベースに基づく最先端コーディングモデルの評価、最高解像度38.8%","summary":"SpecificিকはReal-SWEベンチマークを発表し、実際の企業が承認したプライベート生産コードベースのタスクに対するフロンティアコーディングモデルを評価しました。","category":"業界動向","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWEリリース:民間企業のコードベースに基づく最先端コーディングモデルの評価、最高解像度38.8% - Aioga AIニュース","description":"SpecificিকはReal-SWEベンチマークを発表し、実際の企業が承認したプライベート生産コードベースのタスクに対するフロンティアコーディングモデルを評価しました。","url":"https://www.aioga.com/ja/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:19.806Z"},"ko":{"title":"Real-SWE 릴리스: 민간 기업 코드베이스를 기반으로 한 최첨단 코딩 모델 평가, 최고 해상도 38.8%","summary":"Specificিক은 Real-SWE 벤치마크를 발표했는데, 이는 실제 기업이 승인한 사적 생산 코드베이스의 작업을 위한 프런티어 코딩 모델을 평가했습니다.","category":"업계 동향","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE 릴리스: 민간 기업 코드베이스를 기반으로 한 최첨단 코딩 모델 평가, 최고 해상도 38.8% - Aioga AI 뉴스","description":"Specificিক은 Real-SWE 벤치마크를 발표했는데, 이는 실제 기업이 승인한 사적 생산 코드베이스의 작업을 위한 프런티어 코딩 모델을 평가했습니다.","url":"https://www.aioga.com/ko/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:29.312Z"},"es":{"title":"Real-SWE publicado: Evaluación de modelos de codificación avanzados basados en código privado de empresas, tasa máxima de resolución 38,8%","summary":"Specific lanzó el benchmark Real-SWE, evaluando modelos de codificación avanzados con tareas de código de producción privado autorizadas por empresas reales.","category":"Industria","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE publicado: Evaluación de modelos de codificación avanzados basados en código privado de empresas, tasa máxima de resolución 38,8% - Aioga Noticias de IA","description":"Specific lanzó el benchmark Real-SWE, evaluando modelos de codificación avanzados con tareas de código de producción privado autorizadas por empresas reales.","url":"https://www.aioga.com/es/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:27.519Z"},"fr":{"title":"Releases Real-SWE : Évaluation de modèles de codage de pointe basés sur des bases de code d’entreprises privées, taux de résolution maximal 38,8 %","summary":"Specificিক a publié le benchmark Real-SWE, évaluant des modèles de codage de pointe pour des tâches issues de bases de code de production privées autorisées par de vraies entreprises.","category":"Industrie","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Releases Real-SWE : Évaluation de modèles de codage de pointe basés sur des bases de code d’entreprises privées, taux de résolution maximal 38,8 % - Aioga Actualités IA","description":"Specificিক a publié le benchmark Real-SWE, évaluant des modèles de codage de pointe pour des tâches issues de bases de code de production privées autorisées par de vraies entrepris...","url":"https://www.aioga.com/fr/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:38.968Z"},"de":{"title":"Real-SWE-Veröffentlichungen: Bewertung modernster Codierungsmodelle basierend auf privaten Unternehmenscodebasen, höchste Auflösungsrate 38,8 %","summary":"Specificিক veröffentlichte den Real-SWE-Benchmark, der Frontier-Coding-Modelle für Aufgaben aus privaten Produktionscodebasen bewertete, die von echten Unternehmen autorisiert wurden.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE-Veröffentlichungen: Bewertung modernster Codierungsmodelle basierend auf privaten Unternehmenscodebasen, höchste Auflösungsrate 38,8 % - Aioga KI-News","description":"Specificিক veröffentlichte den Real-SWE-Benchmark, der Frontier-Coding-Modelle für Aufgaben aus privaten Produktionscodebasen bewertete, die von echten Unternehmen autorisiert wurd...","url":"https://www.aioga.com/de/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:38.874Z"},"pt-BR":{"title":"Lançamentos de Real-SWE: Avaliação de Modelos de Codificação de Ponta Baseados em Bases de Código de Empresas Privadas, Taxa de Resolução Máxima 38,8%","summary":"A Specificিক lançou o benchmark Real-SWE, avaliando modelos de codificação de fronteira para tarefas de bases de código de produção privadas autorizadas por empresas reais.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Lançamentos de Real-SWE: Avaliação de Modelos de Codificação de Ponta Baseados em Bases de Código de Empresas Privadas, Taxa de Resolução Máxima 38,8% - Aioga Notícias de IA","description":"A Specificিক lançou o benchmark Real-SWE, avaliando modelos de codificação de fronteira para tarefas de bases de código de produção privadas autorizadas por empresas reais.","url":"https://www.aioga.com/pt-BR/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:48.683Z"},"ru":{"title":"Релизы Real-SWE: оценка передовых моделей кодирования на основе кодовых баз частных предприятий, максимальный уровень разрешения 38,8%","summary":"Specificিক выпустила бенчмарк Real-SWE, оценивающий модели кодирования передов для задач из частных производственных кодовых баз, авторизованных реальными компаниями.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Релизы Real-SWE: оценка передовых моделей кодирования на основе кодовых баз частных предприятий, максимальный уровень разрешения 38,8% - Aioga Новости ИИ","description":"Specificিক выпустила бенчмарк Real-SWE, оценивающий модели кодирования передов для задач из частных производственных кодовых баз, авторизованных реальными компаниями.","url":"https://www.aioga.com/ru/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:48.151Z"},"ar":{"title":"إصدارات Real-SWE: تقييم نماذج الترميز المتطورة بناء على قواعد الشيفرة الخاصة للمؤسسات، أعلى معدل دقة 38.8٪","summary":"أصدرت Specificিক معيار Real-SWE، الذي يقيم نماذج الترميز الحدودية للمهام من قواعد الشيفرة الخاصة المعتمدة من شركات حقيقية.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"إصدارات Real-SWE: تقييم نماذج الترميز المتطورة بناء على قواعد الشيفرة الخاصة للمؤسسات، أعلى معدل دقة 38.8٪ - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت Specificিক معيار Real-SWE، الذي يقيم نماذج الترميز الحدودية للمهام من قواعد الشيفرة الخاصة المعتمدة من شركات حقيقية.","url":"https://www.aioga.com/ar/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:57.881Z"},"hi":{"title":"वास्तविक-SWE रिलीज़: निजी उद्यम कोडबेस के आधार पर अत्याधुनिक कोडिंग मॉडल का मूल्यांकन, उच्चतम रिज़ॉल्यूशन दर 38.8%","summary":"Specificিক ने रियल-एसडब्ल्यूई बेंचमार्क जारी किया, जिसमें वास्तविक कंपनियों द्वारा अधिकृत निजी उत्पादन कोडबेस से कार्यों के लिए फ्रंटियर कोडिंग मॉडल का मूल्यांकन किया गया।","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"वास्तविक-SWE रिलीज़: निजी उद्यम कोडबेस के आधार पर अत्याधुनिक कोडिंग मॉडल का मूल्यांकन, उच्चतम रिज़ॉल्यूशन दर 38.8% - Aioga AI समाचार","description":"Specificিক ने रियल-एसडब्ल्यूई बेंचमार्क जारी किया, जिसमें वास्तविक कंपनियों द्वारा अधिकृत निजी उत्पादन कोडबेस से कार्यों के लिए फ्रंटियर कोडिंग मॉडल का मूल्यांकन किया गया।","url":"https://www.aioga.com/hi/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:01:58.297Z"},"it":{"title":"Real-SWE pubblica: valutazione dei modelli di coding di frontiera basata su repository privati di aziende, tasso massimo di risoluzione 38,8%","summary":"Specific pubblica il benchmark Real-SWE, valutando i modelli di coding di frontiera su compiti tratti da repository di produzione privati autorizzati da aziende reali.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE pubblica: valutazione dei modelli di coding di frontiera basata su repository privati di aziende, tasso massimo di risoluzione 38,8% - Aioga Notizie IA","description":"Specific pubblica il benchmark Real-SWE, valutando i modelli di coding di frontiera su compiti tratti da repository di produzione privati autorizzati da aziende reali.","url":"https://www.aioga.com/it/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:06.285Z"},"nl":{"title":"Real-SWE gepubliceerd: Evaluatie van geavanceerde codeermodellen op basis van privécodebases van bedrijven, hoogste oplossingspercentage 38,8%","summary":"Specific publiceert Real-SWE benchmark, die geavanceerde codeermodellen evalueert met taken afkomstig uit door echte bedrijven goedgekeurde privéproductiecodebases.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE gepubliceerd: Evaluatie van geavanceerde codeermodellen op basis van privécodebases van bedrijven, hoogste oplossingspercentage 38,8% - Aioga AI-nieuws","description":"Specific publiceert Real-SWE benchmark, die geavanceerde codeermodellen evalueert met taken afkomstig uit door echte bedrijven goedgekeurde privéproductiecodebases.","url":"https://www.aioga.com/nl/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:05.631Z"},"tr":{"title":"Real-SWE Sürümleri: Özel Kurumsal Kod Tabanlarına Dayalı En Son Kodlama Modellerinin Değerlendirilmesi, En Yüksek Çözünürlük Oranı %38,8","summary":"Specificিক, gerçek şirketler tarafından yetkilendirilen özel üretim kod tabanlarından gelen görevler için sınır kodlama modellerini değerlendiren Real-SWE kıyaslamasını yayımladı.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE Sürümleri: Özel Kurumsal Kod Tabanlarına Dayalı En Son Kodlama Modellerinin Değerlendirilmesi, En Yüksek Çözünürlük Oranı %38,8 - Aioga AI Haberleri","description":"Specificিক, gerçek şirketler tarafından yetkilendirilen özel üretim kod tabanlarından gelen görevler için sınır kodlama modellerini değerlendiren Real-SWE kıyaslamasını yayımladı.","url":"https://www.aioga.com/tr/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:15.420Z"},"vi":{"title":"Real-SWE Releases: Đánh giá các mô hình mã hóa tiên tiến dựa trên mã nguồn doanh nghiệp tư nhân, tỷ lệ phân giải cao nhất 38,8%","summary":"Specificিক đã phát hành benchmark Real-SWE, đánh giá các mô hình mã hóa biên giới cho các tác vụ từ mã nguồn sản xuất tư nhân được các công ty thực sự cấp phép.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE Releases: Đánh giá các mô hình mã hóa tiên tiến dựa trên mã nguồn doanh nghiệp tư nhân, tỷ lệ phân giải cao nhất 38,8% - Tin tức AI Aioga","description":"Specificিক đã phát hành benchmark Real-SWE, đánh giá các mô hình mã hóa biên giới cho các tác vụ từ mã nguồn sản xuất tư nhân được các công ty thực sự cấp phép.","url":"https://www.aioga.com/vi/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:15.550Z"},"id":{"title":"Rilis Real-SWE: Mengevaluasi Model Pengkodean Mutakhir Berdasarkan Kode Perusahaan Pribadi, Tingkat Resolusi Tertinggi 38,8%","summary":"Specificিক merilis benchmark Real-SWE, mengevaluasi model pengkodean terbatas untuk tugas dari basis kode produksi pribadi yang diotorisasi oleh perusahaan nyata.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Rilis Real-SWE: Mengevaluasi Model Pengkodean Mutakhir Berdasarkan Kode Perusahaan Pribadi, Tingkat Resolusi Tertinggi 38,8% - Berita AI Aioga","description":"Specificিক merilis benchmark Real-SWE, mengevaluasi model pengkodean terbatas untuk tugas dari basis kode produksi pribadi yang diotorisasi oleh perusahaan nyata.","url":"https://www.aioga.com/id/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:25.229Z"},"th":{"title":"Real-SWE Releases: การประเมินโมเดลการเขียนโค้ดล้ําสมัยโดยอิงจากฐานโค้ดขององค์กรเอกชน อัตราความละเอียดสูงสุด 38.8%","summary":"Specificিক ได้ปล่อยมาตรฐาน Real-SWE ซึ่งประเมินโมเดลการเขียนโค้ดแนวหน้าสําหรับงานจากฐานโค้ดผลิตส่วนตัวที่ได้รับอนุญาตจากบริษัทจริง","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Real-SWE Releases: การประเมินโมเดลการเขียนโค้ดล้ําสมัยโดยอิงจากฐานโค้ดขององค์กรเอกชน อัตราความละเอียดสูงสุด 38.8% - ข่าว AI Aioga","description":"Specificিক ได้ปล่อยมาตรฐาน Real-SWE ซึ่งประเมินโมเดลการเขียนโค้ดแนวหน้าสําหรับงานจากฐานโค้ดผลิตส่วนตัวที่ได้รับอนุญาตจากบริษัทจริง","url":"https://www.aioga.com/th/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:24.644Z"},"pl":{"title":"Prawdziwe wydania oprogramowania: Ocena nowoczesnych modeli kodowania opartych na bazach kodu prywatnych przedsiębiorstw, najwyższa rozdzielczość 38,8%","summary":"Specificিক wydało benchmark Real-SOFTVER, oceniający modele kodowania frontier dla zadań z prywatnych baz kodu zatwierdzonych przez prawdziwe firmy.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News：AI 热帖","pageTitle":"Prawdziwe wydania oprogramowania: Ocena nowoczesnych modeli kodowania opartych na bazach kodu prywatnych przedsiębiorstw, najwyższa rozdzielczość 38,8% - Aioga Wiadomości AI","description":"Specificিক wydało benchmark Real-SOFTVER, oceniający modele kodowania frontier dla zadań z prywatnych baz kodu zatwierdzonych przez prawdziwe firmy.","url":"https://www.aioga.com/pl/news/cmtz5c0hd0ujmroup4uua5duy/","contentTranslated":true,"sourceHash":"3e5731e684410eff","translatedAt":"2026-09-13T02:02:34.706Z"}},"evidenceTier":"verified-news","reviewStatus":"automated-ingest","indexable":true,"editorialCover":""}}