{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T07:21:26.498Z","headline":"Feyn AI 发布 SQRL 文本转 SQL 模型族，查询前先探查数据库","description":"Feyn Labs 发布 SQRL 文本转 SQL 模型族，在生成查询前先用只读探针检查数据库。旗舰版 SQRL-35B-A3B 在 BIRD Dev 上达到 70.6% 执行准确率，超越 Claude Opus 4.6。该模型还蒸馏出可自托管的 4B 和 9B 检查点。","url":"https://www.aioga.com/news/cmrsdw0v700mxbiwm7uws4c4z/","mainEntityOfPage":"https://www.aioga.com/news/cmrsdw0v700mxbiwm7uws4c4z/","datePublished":"2026-07-19T22:20:23.000Z","dateModified":"2026-07-19T22:20:23.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.marktechpost.com/2026/07/19/feyn-ai-releases-sqrl-a-text-to-sql-model-family-that-inspects-the-database-before-writing-a-query","https://aihot.virxact.com/items/cmrsdw0v700mxbiwm7uws4c4z"],"canonicalUrl":"https://www.aioga.com/news/cmrsdw0v700mxbiwm7uws4c4z/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：Feyn Labs 发布 SQRL 文本转 SQL 模型族，在生成查询前先用只读探针检查数据库。 Aioga 将其归入「模型更新」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/cmrsdw0v700mxbiwm7uws4c4z/","dateCreated":"2026-07-19T22:20:23.000Z","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":"marktechpost.com source article","url":"https://www.marktechpost.com/2026/07/19/feyn-ai-releases-sqrl-a-text-to-sql-model-family-that-inspects-the-database-before-writing-a-query","datePublished":"2026-07-19T22:20:23.000Z","provider":{"@type":"Organization","name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/19/feyn-ai-releases-sqrl-a-text-to-sql-model-family-that-inspects-the-database-before-writing-a-query"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrsdw0v700mxbiwm7uws4c4z","datePublished":"2026-07-19T22:20:23.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrsdw0v700mxbiwm7uws4c4z"}}],"aggregationSource":"MarkTechPost（RSS）","originalPublisher":{"name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/19/feyn-ai-releases-sqrl-a-text-to-sql-model-family-that-inspects-the-database-before-writing-a-query"},"article":{"id":"cmrsdw0v700mxbiwm7uws4c4z","slug":"cmrsdw0v700mxbiwm7uws4c4z","url":"https://www.aioga.com/news/cmrsdw0v700mxbiwm7uws4c4z/","title":"Feyn AI 发布 SQRL 文本转 SQL 模型族，查询前先探查数据库","title_en":"Feyn AI Releases SQRL， a Text-to-SQL Model Family That Inspects the Database Before Writing a Query","summary":"Feyn Labs 发布 SQRL 文本转 SQL 模型族，在生成查询前先用只读探针检查数据库。旗舰版 SQRL-35B-A3B 在 BIRD Dev 上达到 70.6% 执行准确率，超越 Claude Opus 4.6。该模型还蒸馏出可自托管的 4B 和 9B 检查点。","source":"MarkTechPost（RSS）","sourceUrl":"https://www.marktechpost.com/2026/07/19/feyn-ai-releases-sqrl-a-text-to-sql-model-family-that-inspects-the-database-before-writing-a-query","aiHotUrl":"https://aihot.virxact.com/items/cmrsdw0v700mxbiwm7uws4c4z","publishedAt":"2026-07-19T22:20:23.000Z","category":"模型更新","score":57,"selected":false,"articleBody":["Most text-to-SQL systems treat the task as translation. Feyn AI：https://usefeyn.com/ (YC-backed startup) reframes it around inspection. The Feyn team has released SQRL：https://usefeyn.com/blog/sqrl-asks-the-database-first/, a family of models that turn natural language questions into SQL. Instead of generating a query immediately, SQRL can inspect the database first. This lets it resolve ambiguity and write only queries the data actually supports.","Feyn team reports that the flagship SQRL-35B-A3B reaches 70.6% execution accuracy on BIRD：https://bird-bench.github.io/ Dev . That figure edges Claude Opus 4.6 at 68.77% under the same evaluation. Three checkpoints ship openly on Hugging Face：https://huggingface.co/feyninc: SQRL-4B：https://huggingface.co/feyninc/sqrl-4b, SQRL-9B：https://huggingface.co/feyninc/sqrl-9b, and SQRL-35B-A3B：https://huggingface.co/feyninc/sqrl-35b-a3b.","Text-to-SQL is often described as a translation problem, but that framing misses the hardest part. A query can be perfectly valid SQL and still return the wrong answer. It can join the wrong tables, read an ambiguous column incorrectly, or filter for values that do not exist. None of these mistakes throws an error, so none is caught by execution alone.","Schema information does not prevent them. A schema lists tables, columns, types, and sometimes relationships. It does not reveal whether a county is stored as Alameda , Alameda County , or ALAMEDA . It cannot tell you which join produces duplicate rows.","The BIRD benchmark：https://arxiv.org/abs/2305.03111 makes these failures measurable. Its databases span real domains and contain imperfect values, ambiguous columns, and nontrivial relationships. A system is scored by executing its SQL and comparing the returned rows against a reference result. For query languages, syntactic correctness is not enough. Feyn’s core insight is that the missing information already lives inside the database. The model simply needs permission to ask for it.","SQRL receives a question, its schema, and optional evidence about the database. If that context is enough, it returns a query at once. If something remains ambiguous, it runs read-only queries and uses the returned rows to draft its final answer. The decision to inspect is circumstantial. Counting rows in a single table needs no lookup, so SQRL answers directly.","The interaction uses two distinct actions. An block requests an observation from the database. An block commits to the final query. The harness executes exploration queries in read-only mode and returns their rows inside tags. SQRL can inspect up to five times, though most questions finish in fewer steps.","The explainer below walks through both behaviors on real examples, then compares the family against frontier models on BIRD Dev.","Text-to-SQL has historically followed two approaches, and each trades away something important.","Single-shot models generate the entire query in one turn. They are inexpensive to serve, but they must infer everything from the schema, which can produce logically incorrect queries. Frontier pipelines instead retrieve context, generate candidates, critique them, and then select an answer. This can maximize accuracy, but every question requires several expensive frontier calls and multiple database round trips. That cost makes such pipelines hard to place on the hot path.","SQRL combines both in a single model. Easy questions stay short, while ambiguous questions earn an inspection. The model pays the cost of looking only when the question needs it.","Giving a model database access does not teach it when or how to look. That behavior has to be trained, and execution-based training is unforgiving. If a reference query is itself wrong, a correct model answer receives the wrong reward. Feyn therefore cleaned the training pool first. Starting from BIRD and Spider：https://arxiv.org/abs/1809.08887, it removed examples whose reference SQL produced no usable result. Three model judges then reviewed the remaining pairs and dropped any query that did not answer its question as written. The test split combines a held-out Spider split with BIRD dev, and the rest of the data went into training.","The 35B-A3B teacher trained directly with CISPO , a reinforcement learning method from MiniMax’s M1 work：https://arxiv.org/abs/2506.13585. CISPO clips the importance-sampling weights rather than the policy ratio, which preserves gradient signal from rare but decisive tokens. For each question, the model produced eight complete trajectories. Feyn executed every final query and rewarded a result that matched the reference, a binary signal that ignores wording and checks only the returned rows.","Group-relative training needs variation inside each group. Eight correct or eight failed attempts carry no signal about which decisions helped. Feyn therefore trained on the ‘mixed zone,’ where only some of the eight attempts succeeded, so every group could reinforce the choices that separated a correct trajectory from an incorrect one. That is how the teacher learned when to inspect.","To make the behavior deployable, Feyn team sampled complete teacher trajectories and kept only runs whose final SQL returned the correct result. This produced about 10,200 examples, each preserving the reasoning, exploration queries, observations, and final answer. The 4B and 9B students were fine-tuned on these trajectories, then refined with the same CISPO execution reward. SQRL builds on the Qwen3.5 and Qwen3.6 model families.","Feyn evaluated SQRL on BIRD Dev, scoring a query correct when it returns the same result as the reference. SQRL-35B-A3B scores 70.60% and activates about 3B parameters per token. The 9B student holds nearly all of that at 69.80%. SQRL-4B reaches 68.80%, matching Claude Opus 4.6 on this evaluation in a model small enough to host anywhere, so your schema, queries, and observations stay on infrastructure you control.","In Feyn’s reported comparison, the frontier field trails: Claude 4.5 Sonnet at 67.34%, Qwen3-Coder-480B-A35B at 66.17%, GLM-4.7 at 63.82%, DeepSeek-R1 at 61.67%, and Kimi-K2-Thinking at 60.63%.","Feyn recommends SQRL-9B as the default checkpoint, with SQRL-4B for the tightest budgets and SQRL-35B-A3B for the highest accuracy. The 9B model serves with vLLM：https://github.com/vllm-project/vllm:","The application loop is small. Keep database execution read-only, and return each observation to the model until it emits an answer. One caveat matters. Do not enable a serving-layer reasoning parser. The action protocol appears in the content after the closing tag, so stripping that content removes the model’s or action. Parse the raw message content and preserve everything after the final think tag. The model cards contain the complete system prompt and a reference harness.","Asif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences.","Build an Agentic Event Venue Operator [Full Codes]：https://pxllnk.co/twdn5","Thanks! Our team will contact you soon 🙌"],"articleImages":[{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2019/06/Screen-Shot-2021-09-14-at-9.02.24-AM-300x300.png","alt":"","afterParagraph":18,"url":"/media/articles/cmrsdw0v700mxbiwm7uws4c4z/787a6d54564e8e19.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/blog19132-1-8-100x70.png","alt":"Google Releases Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber","afterParagraph":19,"url":"/media/articles/cmrsdw0v700mxbiwm7uws4c4z/39896896ab8c8014.webp"}],"mediaStatus":"ok","articleBodyZh":["大多数文本到 SQL 系统将此任务视为翻译。Feyn AI：https://usefeyn.com/（YC 支持的初创公司）将其重新定义为围绕检查。Feyn 团队发布了 SQRL：https://usefeyn.com/blog/sqrl-asks-the-database-first/，这是一个将自然语言问题转换为 SQL 的模型系列。SQRL 并不是立即生成查询，而是可以先检查数据库。这使它能够解决歧义，并只编写数据实际支持的查询。","Feyn 团队报告称，旗舰模型 SQRL-35B-A3B 在 BIRD：https://bird-bench.github.io/ Dev 上达到 70.6% 的执行准确率。这个数字在相同评估下略高于 Claude Opus 4.6 的 68.77%。三个检查点在 Hugging Face：https://huggingface.co/feyninc 上公开发布：SQRL-4B：https://huggingface.co/feyninc/sqrl-4b，SQRL-9B：https://huggingface.co/feyninc/sqrl-9b，以及 SQRL-35B-A3B：https://huggingface.co/feyninc/sqrl-35b-a3b。","文本到 SQL 通常被描述为一个翻译问题，但这种表述忽略了最困难的部分。查询可能是完全有效的 SQL，却仍然返回错误的答案。它可能连接错误的表，错误地读取模糊的列，或者过滤不存在的值。所有这些错误都不会抛出异常，因此仅通过执行无法发现。","模式信息也不能防止这些错误。一个模式列出了表、列、类型，有时还有关系。但它无法显示一个县是以 Alameda、Alameda County 还是 ALAMEDA 存储。它也无法告诉你哪个连接会生成重复行。","BIRD 基准：https://arxiv.org/abs/2305.03111 使这些失败可以量化。其数据库涵盖真实领域，包含不完善的值、模糊列和复杂关系。一个系统的评分方式是执行其 SQL 并将返回的行与参考结果进行比较。对于查询语言来说，语法正确是不够的。Feyn 的核心洞见是，缺失的信息已经存在于数据库中。模型只需要获得请求这些信息的权限。","SQRL 接收一个问题、其模式以及可选的数据库证据。如果这些上下文足够，它会立即返回查询。如果还有不明确的地方，它会运行只读查询，并利用返回的行来起草最终答案。是否进行检查取决于情况。对单个表计数行不需要查找，因此 SQRL 会直接回答。","交互使用两种不同的操作。一个块请求数据库的观察结果。一个块提交最终查询。运行环境以只读模式执行探索查询，并将返回的行放在标签内。SQRL 最多可以检查五次，尽管大多数问题在更少步骤内完成。","下面的讲解通过真实示例说明了这两种行为，然后将该家族模型与 BIRD Dev 上的前沿模型进行比较。","Text-to-SQL 历来有两种方法，每种方法都要放弃一些重要的东西。","一次性模型在一次操作中生成整个查询。它们服务成本低，但必须从模式中推断所有内容，这可能产生逻辑上不正确的查询。前沿流水线则先检索上下文，生成候选项，进行批评，然后选择答案。这可以最大化准确性，但每个问题都需要多次昂贵的前沿调用和多次数据库往返。这种成本使得此类流水线难以应用于关键路径。","SQRL 将两者结合在单一模型中。简单问题保持简短，模糊问题则进行检查。模型仅在问题需要时才支付检查的成本。","给模型数据库访问权限并不会教它何时或如何查找。该行为必须经过训练，而基于执行的训练是无情的。如果参考查询本身错误，正确的模型答案将获得错误的奖励。因此 Feyn 首先清理了训练数据。以 BIRD 和 Spider：https://arxiv.org/abs/1809.08887 为起点，它删除了参考 SQL 没有产生可用结果的示例。然后由三位模型评审人员审核剩余的配对，并删除任何未按照原问题回答的查询。测试集结合了保留的 Spider 划分和 BIRD 开发集，其余数据用于训练。","35B-A3B 教师模型直接使用 CISPO 进行训练，CISPO 是 MiniMax 的 M1 工作中的一种强化学习方法：https://arxiv.org/abs/2506.13585。CISPO 截断重要性采样权重，而不是策略比率，这保留了来自罕见但决定性 token 的梯度信号。对于每个问题，模型生成了八条完整的轨迹。Feyn 执行每个最终查询，并对匹配参考结果的结果给予奖励，这是一种二值信号，仅检查返回的行，而忽略措辞。","群组相对训练需要每个群组内部的变化。八次正确或八次失败的尝试无法提供哪些决策有帮助的信号。因此 Feyn 在“混合区”上进行训练，其中八次尝试中只有部分成功，从而每个群组都可以强化区分正确轨迹与错误轨迹的选择。教师就是这样学会何时进行检查的。","为了使行为可部署，Feyn 团队采样完整的教师轨迹，并仅保留最终 SQL 返回正确结果的运行。这产生了约 10,200 个示例，每个示例保留了推理、探索查询、观察和最终答案。4B 和 9B 学生模型在这些轨迹上进行了微调，然后用相同的 CISPO 执行奖励进行了细化。SQRL 基于 Qwen3.5 和 Qwen3.6 模型系列。","Feyn 在 BIRD Dev 上评估了 SQRL，当查询返回的结果与参考结果一致时，判定为正确。SQRL-35B-A3B 得分 70.60%，每个 token 激活约 30 亿参数。9B 学生模型保留了几乎全部能力，得分为 69.80%。SQRL-4B 达到 68.80%，在此评估中与 Claude Opus 4.6 匹配，并且模型足够小，可在任意位置托管，从而让您的架构、查询和观察保持在您控制的基础设施上。","在 Feyn 报告的比较中，前沿领域的其他模型表现稍差：Claude 4.5 Sonnet 为 67.34%，Qwen3-Coder-480B-A35B 为 66.17%，GLM-4.7 为 63.82%，DeepSeek-R1 为 61.67%，Kimi-K2-Thinking 为 60.63%。","Feyn 推荐 SQRL-9B 作为默认检查点，预算最紧时使用 SQRL-4B，追求最高准确率时使用 SQRL-35B-A3B。9B 模型可使用 vLLM 服务：https://github.com/vllm-project/vllm","应用程序循环很小。保持数据库执行为只读，并将每个观察结果返回给模型，直到它发出答案。有一个注意事项很重要。不要启用服务层推理解析器。操作协议出现在结束标签之后的内容中，因此删除该内容会移除模型或操作。解析原始消息内容并保留最终 think 标签之后的所有内容。模型卡包含完整的系统提示和参考工具。","Asif Razzaq 是 Marktechpost Media Inc. 的首席执行官。作为一名有远见的企业家和工程师，Asif 致力于利用人工智能的潜力造福社会。他最近的努力是推出人工智能媒体平台 Marktechpost，该平台以深入报道机器学习和深度学习资讯而闻名，内容既具技术性，又易于广大观众理解。该平台每月浏览量超过 200 万次，显示了其在观众中的受欢迎程度。","构建智能活动场地运营商 [完整代码]：https://pxllnk.co/twdn5","谢谢！我们的团队会尽快联系您 🙌"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：Feyn Labs 发布 SQRL 文本转 SQL 模型族，在生成查询前先用只读探针检查数据库。 Aioga 将其归入「模型更新」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":"2026-07-23T07:30:40.541Z","sourceHash":"3cc3de7bf58a6960","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["模型更新","MarkTechPost（RSS）"],"translations":{"zh-CN":{"title":"Feyn AI 发布 SQRL 文本转 SQL 模型族，查询前先探查数据库","summary":"Feyn Labs 发布 SQRL 文本转 SQL 模型族，在生成查询前先用只读探针检查数据库。旗舰版 SQRL-35B-A3B 在 BIRD Dev 上达到 70.6% 执行准确率，超越 Claude Opus 4.6。该模型还蒸馏出可自托管的 4B 和 9B 检查点。","category":"模型更新","source":"marktechpost.com","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI 发布 SQRL 文本转 SQL 模型族，查询前先探查数据库 - Aioga AI资讯","description":"Feyn Labs 发布 SQRL 文本转 SQL 模型族，在生成查询前先用只读探针检查数据库。旗舰版 SQRL-35B-A3B 在 BIRD Dev 上达到 70.6% 执行准确率，超越 Claude Opus 4.6。该模型还蒸馏出可自托管的 4B 和 9B 检查点。","url":"https://www.aioga.com/news/cmrsdw0v700mxbiwm7uws4c4z/"},"en":{"title":"Feyn AI releases the SQRL text-to-SQL model family, allowing database exploration before querying","summary":"Feyn Labs released the SQRL text-to-SQL model family, which uses a read-only probe to check the database before generating queries. The flagship SQRL-35B-A3B achieved 70.6% execution accuracy on BIRD Dev, surpassing Claude Opus 4.6. The model also distilled self-hosted 4B and 9B checkpoints.","category":"Models","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI releases the SQRL text-to-SQL model family, allowing database exploration before querying - Aioga AI News","description":"Feyn Labs released the SQRL text-to-SQL model family, which uses a read-only probe to check the database before generating queries. The flagship SQRL-35B-A3B achieved 70.6% executi...","url":"https://www.aioga.com/en/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:41:49.588Z"},"ja":{"title":"Feyn AI、SQRLテキストからSQLへのモデルファミリーを発表、クエリの前にデータベースを検査","summary":"Feyn LabsはSQRLテキストからSQLへのモデルファミリーを公開し、クエリを生成する前に読み取り専用プローブでデータベースをチェックします。フラッグシップモデルSQRL-35B-A3BはBIRD Devで70.6％の実行精度を達成し、Claude Opus 4.6を上回りました。このモデルは、自己ホスト可能な4Bおよび9Bのチェックポイントも蒸留しています。","category":"モデル更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI、SQRLテキストからSQLへのモデルファミリーを発表、クエリの前にデータベースを検査 - Aioga AIニュース","description":"Feyn LabsはSQRLテキストからSQLへのモデルファミリーを公開し、クエリを生成する前に読み取り専用プローブでデータベースをチェックします。フラッグシップモデルSQRL-35B-A3BはBIRD Devで70.6％の実行精度を達成し、Claude Opus 4.6を上回りました。このモデルは、自己ホスト可能な4Bおよび9Bのチェックポイントも蒸留して...","url":"https://www.aioga.com/ja/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:41:57.039Z"},"ko":{"title":"Feyn AI가 SQRL 텍스트를 SQL로 변환하는 모델 패밀리를 출시, 쿼리 전에 먼저 데이터베이스 탐색","summary":"Feyn Labs는 SQRL 텍스트를 SQL로 변환하는 모델군을 발표했으며, 쿼리를 생성하기 전에 읽기 전용 프로브로 데이터베이스를 검사합니다. 플래그십 모델 SQRL-35B-A3B는 BIRD Dev에서 70.6% 실행 정확도를 달성하여 Claude Opus 4.6을 능가했습니다. 이 모델은 또한 자체 호스팅 가능한 4B 및 9B 체크포인트로 증류되었습니다.","category":"모델 업데이트","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI가 SQRL 텍스트를 SQL로 변환하는 모델 패밀리를 출시, 쿼리 전에 먼저 데이터베이스 탐색 - Aioga AI 뉴스","description":"Feyn Labs는 SQRL 텍스트를 SQL로 변환하는 모델군을 발표했으며, 쿼리를 생성하기 전에 읽기 전용 프로브로 데이터베이스를 검사합니다. 플래그십 모델 SQRL-35B-A3B는 BIRD Dev에서 70.6% 실행 정확도를 달성하여 Claude Opus 4.6을 능가했습니다. 이 모델은 또한 자체 호스팅 가능한 4B...","url":"https://www.aioga.com/ko/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:42:42.453Z"},"es":{"title":"Feyn AI lanza la familia de modelos SQRL para convertir texto a SQL, explorando la base de datos antes de la consulta","summary":"Feyn Labs lanzó la familia de modelos SQRL para convertir texto a SQL, utilizando una sonda de solo lectura para verificar la base de datos antes de generar consultas. La versión insignia SQRL-35B-A3B alcanzó una precisión de ejecución del 70,6 % en BIRD Dev, superando a Claude Opus 4.6. Este modelo también destiló puntos de control autohospedables de 4B y 9B.","category":"Modelos","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI lanza la familia de modelos SQRL para convertir texto a SQL, explorando la base de datos antes de la consulta - Aioga Noticias de IA","description":"Feyn Labs lanzó la familia de modelos SQRL para convertir texto a SQL, utilizando una sonda de solo lectura para verificar la base de datos antes de generar consultas. La versión i...","url":"https://www.aioga.com/es/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:42:36.795Z"},"fr":{"title":"Feyn AI lance la famille de modèles SQRL de conversion de texte en SQL, explorant d'abord la base de données avant la requête","summary":"Feyn Labs a publié la famille de modèles SQRL de conversion de texte en SQL, qui vérifie d'abord la base de données avec une sonde en lecture seule avant de générer des requêtes. La version phare SQRL-35B-A3B a atteint un taux de précision d'exécution de 70,6 % sur BIRD Dev, dépassant Claude Opus 4.6. Ce modèle a également distillé des points de contrôle autohébergés de 4B et 9B.","category":"Modèles","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI lance la famille de modèles SQRL de conversion de texte en SQL, explorant d'abord la base de données avant la requête - Aioga Actualités IA","description":"Feyn Labs a publié la famille de modèles SQRL de conversion de texte en SQL, qui vérifie d'abord la base de données avec une sonde en lecture seule avant de générer des requêtes. L...","url":"https://www.aioga.com/fr/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:43:24.305Z"},"de":{"title":"Feyn AI veröffentlicht die SQRL-Text-zu-SQL-Modellreihe, um die Datenbank vor der Abfrage zu erkunden","summary":"Feyn Labs veröffentlicht die SQRL Text-zu-SQL-Modellreihe, die vor der Generierung von Abfragen die Datenbank mit einem reinen Lesesonde überprüft. Die Flaggschiff-Version SQRL-35B-A3B erreicht auf BIRD Dev eine Ausführungsgenauigkeit von 70,6 % und übertrifft damit Claude Opus 4.6. Das Modell destilliert außerdem selbst hostbare 4B- und 9B-Checkpoints.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI veröffentlicht die SQRL-Text-zu-SQL-Modellreihe, um die Datenbank vor der Abfrage zu erkunden - Aioga KI-News","description":"Feyn Labs veröffentlicht die SQRL Text-zu-SQL-Modellreihe, die vor der Generierung von Abfragen die Datenbank mit einem reinen Lesesonde überprüft. Die Flaggschiff-Version SQRL-35B...","url":"https://www.aioga.com/de/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:43:20.829Z"},"pt-BR":{"title":"Feyn AI lança a família de modelos SQRL de conversão de texto para SQL, explorando o banco de dados antes da consulta","summary":"A Feyn Labs lançou a família de modelos SQRL de conversão de texto para SQL, que verifica o banco de dados com uma sonda somente leitura antes de gerar consultas. A versão principal SQRL-35B-A3B alcançou 70,6% de precisão de execução no BIRD Dev, superando o Claude Opus 4.6. Esse modelo também destilou pontos de verificação autossustentáveis de 4B e 9B.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI lança a família de modelos SQRL de conversão de texto para SQL, explorando o banco de dados antes da consulta - Aioga Notícias de IA","description":"A Feyn Labs lançou a família de modelos SQRL de conversão de texto para SQL, que verifica o banco de dados com uma sonda somente leitura antes de gerar consultas. A versão principa...","url":"https://www.aioga.com/pt-BR/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:44:05.635Z"},"ru":{"title":"Feyn AI выпустила семейство моделей SQRL для преобразования текста в SQL, сначала исследуя базу данных перед запросом","summary":"Feyn Labs выпустила семейство моделей SQRL для преобразования текста в SQL, предварительно проверяя базу данных с помощью только для чтения зонда перед генерацией запросов. Флагманская версия SQRL-35B-A3B достигла 70,6% точности выполнения на BIRD Dev, превзойдя Claude Opus 4.6. Эта модель также была сжата в самоуправляемые контрольные точки 4B и 9B.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI выпустила семейство моделей SQRL для преобразования текста в SQL, сначала исследуя базу данных перед запросом - Aioga Новости ИИ","description":"Feyn Labs выпустила семейство моделей SQRL для преобразования текста в SQL, предварительно проверяя базу данных с помощью только для чтения зонда перед генерацией запросов. Флагман...","url":"https://www.aioga.com/ru/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:44:06.031Z"},"ar":{"title":"أصدرت Feyn AI عائلة نماذج SQRL لتحويل النصوص إلى SQL، للتحقق من قاعدة البيانات قبل الاستعلام","summary":"أصدرت Feyn Labs مجموعة نماذج تحويل النص إلى SQL SQRL، حيث تقوم أولاً بفحص قاعدة البيانات باستخدام مسبار للقراءة فقط قبل توليد الاستعلامات. النسخة الرائدة SQRL-35B-A3B حققت دقة تنفيذ بلغت 70.6٪ على BIRD Dev، متجاوزة Claude Opus 4.6. كما قام هذا النموذج بتقطير نقاط تفتيش قابلة للاستضافة الذاتية بسعة 4B و9B.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"أصدرت Feyn AI عائلة نماذج SQRL لتحويل النصوص إلى SQL، للتحقق من قاعدة البيانات قبل الاستعلام - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت Feyn Labs مجموعة نماذج تحويل النص إلى SQL SQRL، حيث تقوم أولاً بفحص قاعدة البيانات باستخدام مسبار للقراءة فقط قبل توليد الاستعلامات. النسخة الرائدة SQRL-35B-A3B حققت دقة تنفي...","url":"https://www.aioga.com/ar/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:44:49.595Z"},"hi":{"title":"Feyn AI ने SQRL टेक्स्ट-से-SQL मॉडल श्रृंखला जारी की, क्वेरी करने से पहले डेटाबेस का पता लगाएं","summary":"Feyn Labs ने SQRL टेक्स्ट-टू-SQL मॉडल परिवार जारी किया, जो क्वेरी जनरेट करने से पहले डेटाबेस को सिर्फ पढ़ने वाले प्रॉब के साथ जांचता है। फ्लैगशिप संस्करण SQRL-35B-A3B ने BIRD Dev पर 70.6% निष्पादन सटीकता प्राप्त की, Claude Opus 4.6 को पीछे छोड़ दिया। इस मॉडल ने स्वयं होस्ट किए जाने योग्य 4B और 9B चेकपॉइंट भी डिस्टिल किए।","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI ने SQRL टेक्स्ट-से-SQL मॉडल श्रृंखला जारी की, क्वेरी करने से पहले डेटाबेस का पता लगाएं - Aioga AI समाचार","description":"Feyn Labs ने SQRL टेक्स्ट-टू-SQL मॉडल परिवार जारी किया, जो क्वेरी जनरेट करने से पहले डेटाबेस को सिर्फ पढ़ने वाले प्रॉब के साथ जांचता है। फ्लैगशिप संस्करण SQRL-35B-A3B ने BIRD Dev प...","url":"https://www.aioga.com/hi/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:44:52.509Z"},"it":{"title":"Feyn AI rilascia la famiglia di modelli SQRL per la conversione di testo in SQL, esplorando il database prima della query","summary":"Feyn Labs ha lanciato la famiglia di modelli SQRL per la conversione di testo in SQL, che utilizza una sonda in sola lettura per controllare il database prima di generare le query. La versione di punta SQRL-35B-A3B ha raggiunto un'accuratezza di esecuzione del 70,6% su BIRD Dev, superando Claude Opus 4.6. Il modello ha anche distillato checkpoint auto-ospitabili da 4B e 9B.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI rilascia la famiglia di modelli SQRL per la conversione di testo in SQL, esplorando il database prima della query - Aioga Notizie IA","description":"Feyn Labs ha lanciato la famiglia di modelli SQRL per la conversione di testo in SQL, che utilizza una sonda in sola lettura per controllare il database prima di generare le query....","url":"https://www.aioga.com/it/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:45:34.938Z"},"nl":{"title":"Feyn AI lanceert de SQRL-tekstsnaar-naar-SQL modellenfamilie, eerst de database verkennen voordat je queryt","summary":"Feyn Labs heeft de SQRL-tekst-naar-SQL modelreeks uitgebracht, die de database eerst controleert met een alleen-lezen probe voordat queries worden gegenereerd. Het vlaggenschip SQRL-35B-A3B bereikte een uitvoeringsnauwkeurigheid van 70,6% op BIRD Dev, waarmee het Claude Opus 4.6 overtrof. Het model distilleerde ook zelf-hostbare 4B- en 9B-checkpoints.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI lanceert de SQRL-tekstsnaar-naar-SQL modellenfamilie, eerst de database verkennen voordat je queryt - Aioga AI-nieuws","description":"Feyn Labs heeft de SQRL-tekst-naar-SQL modelreeks uitgebracht, die de database eerst controleert met een alleen-lezen probe voordat queries worden gegenereerd. Het vlaggenschip SQR...","url":"https://www.aioga.com/nl/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:45:32.288Z"},"tr":{"title":"Feyn AI, SQRL metni SQL model ailesini yayınladı, sorgulamadan önce veritabanını keşfedin","summary":"Feyn Labs, SQRL metin tabanlı SQL model ailesini yayınladı ve sorgu üretmeden önce veritabanını yalnızca okunabilir problarla kontrol ediyor. Amiral gemisi SQRL-35B-A3B, BIRD Dev üzerinde %70,6 yürütme doğruluğuna ulaştı ve Claude Opus 4.6'yı geride bıraktı. Bu model ayrıca kendi kendine barındırılabilen 4B ve 9B kontrol noktalarını da damıttı.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI, SQRL metni SQL model ailesini yayınladı, sorgulamadan önce veritabanını keşfedin - Aioga AI Haberleri","description":"Feyn Labs, SQRL metin tabanlı SQL model ailesini yayınladı ve sorgu üretmeden önce veritabanını yalnızca okunabilir problarla kontrol ediyor. Amiral gemisi SQRL-35B-A3B, BIRD Dev ü...","url":"https://www.aioga.com/tr/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:46:14.981Z"},"vi":{"title":"Feyn AI ra mắt dòng mô hình SQRL chuyển đổi văn bản sang SQL, kiểm tra cơ sở dữ liệu trước khi truy vấn","summary":"Feyn Labs phát hành họ mô hình SQRL chuyển đổi văn bản sang SQL, dùng probe chỉ đọc để kiểm tra cơ sở dữ liệu trước khi tạo truy vấn. Phiên bản cao cấp SQRL-35B-A3B đạt độ chính xác thực thi 70,6% trên BIRD Dev, vượt qua Claude Opus 4.6. Mô hình này còn chiết xuất ra các checkpoint 4B và 9B có thể tự lưu trữ.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI ra mắt dòng mô hình SQRL chuyển đổi văn bản sang SQL, kiểm tra cơ sở dữ liệu trước khi truy vấn - Tin tức AI Aioga","description":"Feyn Labs phát hành họ mô hình SQRL chuyển đổi văn bản sang SQL, dùng probe chỉ đọc để kiểm tra cơ sở dữ liệu trước khi tạo truy vấn. Phiên bản cao cấp SQRL-35B-A3B đạt độ chính xá...","url":"https://www.aioga.com/vi/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:46:15.608Z"},"id":{"title":"Feyn AI meluncurkan keluarga model SQRL untuk mengubah teks menjadi SQL, memeriksa database sebelum melakukan query","summary":"Feyn Labs merilis keluarga model SQRL untuk mengubah teks menjadi SQL, menggunakan probe hanya-baca untuk memeriksa basis data sebelum menghasilkan kueri. Versi andalan SQRL-35B-A3B mencapai akurasi eksekusi 70,6% pada BIRD Dev, melampaui Claude Opus 4,6. Model ini juga menurunkan versi checkpoint 4B dan 9B yang bisa di-host sendiri.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI meluncurkan keluarga model SQRL untuk mengubah teks menjadi SQL, memeriksa database sebelum melakukan query - Berita AI Aioga","description":"Feyn Labs merilis keluarga model SQRL untuk mengubah teks menjadi SQL, menggunakan probe hanya-baca untuk memeriksa basis data sebelum menghasilkan kueri. Versi andalan SQRL-35B-A3...","url":"https://www.aioga.com/id/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:47:03.251Z"},"th":{"title":"Feyn AI เปิดตัวชุดโมเดล SQRL สำหรับแปลงข้อความเป็น SQL, สำรวจฐานข้อมูลก่อนการค้นหา","summary":"Feyn Labs เปิดตัวชุดโมเดล SQRL สำหรับแปลงข้อความเป็น SQL ซึ่งก่อนสร้างคำสั่งค้นหาจะใช้โพรบแบบอ่านอย่างเดียวเพื่อตรวจสอบฐานข้อมูล รุ่นแฟลกชิป SQRL-35B-A3B ทำความแม่นยำในการประมวลผลได้ 70.6% บน BIRD Dev เหนือกว่า Claude Opus 4.6 โมเดลนี้ยังสกัดรุ่น 4B และ 9B ที่สามารถโฮสต์เองได้ด้วย","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI เปิดตัวชุดโมเดล SQRL สำหรับแปลงข้อความเป็น SQL, สำรวจฐานข้อมูลก่อนการค้นหา - ข่าว AI Aioga","description":"Feyn Labs เปิดตัวชุดโมเดล SQRL สำหรับแปลงข้อความเป็น SQL ซึ่งก่อนสร้างคำสั่งค้นหาจะใช้โพรบแบบอ่านอย่างเดียวเพื่อตรวจสอบฐานข้อมูล รุ่นแฟลกชิป SQRL-35B-A3B ทำความแม่นยำในการประมวลผลไ...","url":"https://www.aioga.com/th/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:47:07.536Z"},"pl":{"title":"Feyn AI wypuściło rodzinę modeli SQRL do konwersji tekstu na SQL, umożliwiając najpierw eksplorację bazy danych przed zapytaniem","summary":"Feyn Labs wydaje serię modeli SQRL do konwersji tekstu na SQL, które przed wygenerowaniem zapytania sprawdzają bazę danych za pomocą sondy tylko do odczytu. Flagowy model SQRL-35B-A3B osiągnął 70,6% dokładności wykonania na BIRD Dev, przewyższając Claude Opus 4.6. Model ten został również zdestylowany do samodzielnie hostowanych punktów kontrolnych 4B i 9B.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Feyn AI wypuściło rodzinę modeli SQRL do konwersji tekstu na SQL, umożliwiając najpierw eksplorację bazy danych przed zapytaniem - Aioga Wiadomości AI","description":"Feyn Labs wydaje serię modeli SQRL do konwersji tekstu na SQL, które przed wygenerowaniem zapytania sprawdzają bazę danych za pomocą sondy tylko do odczytu. Flagowy model SQRL-35B-...","url":"https://www.aioga.com/pl/news/cmrsdw0v700mxbiwm7uws4c4z/","contentTranslated":true,"sourceHash":"a07542a3ff6514db","translatedAt":"2026-07-23T00:47:53.463Z"}}}}