{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-09-18T17:00:43.709Z","headline":"Anthropic：Claude 四周优化 30 多个开源生物分子模型，平均提速约 4 倍并开源代码","description":"Anthropic 让 Claude 在不到四周内优化了超过 30 个开源生物分子模型，平均提速约 4 倍、精度损失极小，输出完全一致时也有近 2 倍加速。优化代码全部开源，还推出低内存 Big 模式，可在单张 NVIDIA GPU 节点上对超过 10，000 token 的生物分子系统做准确预测。","url":"https://www.aioga.com/news/cmu5y1dkt069qroiqnnhxz778/","mainEntityOfPage":"https://www.aioga.com/news/cmu5y1dkt069qroiqnnhxz778/","datePublished":"2026-09-16T16:00:00.000Z","dateModified":"2026-09-16T16:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.anthropic.com/research/claude-uplifts-biomolecular-modeling","https://aihot.news/items/cmu5y1dkt069qroiqnnhxz778"],"canonicalUrl":"https://www.aioga.com/news/cmu5y1dkt069qroiqnnhxz778/","directAnswer":{"@type":"Answer","text":"Anthropic表示，Claude在不到四周内优化了30多个开源生物分子模型，平均提速约4倍；在输出完全一致时接近2倍，并开源全部优化代码。","url":"https://www.aioga.com/news/cmu5y1dkt069qroiqnnhxz778/","dateCreated":"2026-09-16T16:00:00.000Z","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":"Anthropic source article","url":"https://www.anthropic.com/research/claude-uplifts-biomolecular-modeling","datePublished":"2026-09-16T16:00:00.000Z","provider":{"@type":"Organization","name":"Anthropic","url":"https://www.anthropic.com/research/claude-uplifts-biomolecular-modeling"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/cmu5y1dkt069qroiqnnhxz778","datePublished":"2026-09-16T16:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/cmu5y1dkt069qroiqnnhxz778"}}],"aggregationSource":"Anthropic：Research（发表成果 · 网页）","originalPublisher":{"name":"Anthropic","url":"https://www.anthropic.com/research/claude-uplifts-biomolecular-modeling"},"geoDeepAnswer":null,"article":{"id":"cmu5y1dkt069qroiqnnhxz778","slug":"cmu5y1dkt069qroiqnnhxz778","url":"https://www.aioga.com/news/cmu5y1dkt069qroiqnnhxz778/","title":"Anthropic：Claude 四周优化 30 多个开源生物分子模型，平均提速约 4 倍并开源代码","title_en":"","summary":"Anthropic 让 Claude 在不到四周内优化了超过 30 个开源生物分子模型，平均提速约 4 倍、精度损失极小，输出完全一致时也有近 2 倍加速。优化代码全部开源，还推出低内存 Big 模式，可在单张 NVIDIA GPU 节点上对超过 10，000 token 的生物分子系统做准确预测。","source":"Anthropic：Research（发表成果 · 网页）","sourceUrl":"https://www.anthropic.com/research/claude-uplifts-biomolecular-modeling","aiHotUrl":"https://aihot.news/items/cmu5y1dkt069qroiqnnhxz778","publishedAt":"2026-09-16T16:00:00.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["In this post, we share how Claude made the open-source models that scientists use to predict and design biomolecules faster and more memory-efficient. Claude optimized more than 30 of these models in just under four weeks, speeding them up roughly 4x on average. It also created a low-memory mode that enables the accurate prediction of biomolecular systems larger than 10,000 tokens (amino acids, nucleotides, and atoms from small molecules and ions) on a single NVIDIA GPU node. We are open-sourcing all of the optimized code and announcing a protein design competition co-sponsored with Adaptyv Bio, backed by up to $1 million in Claude credits and wet lab validation for over 5,000 designs. Recently, we shared results：https://www.anthropic.com/research/Claude-accelerates-protein-design demonstrating Claude’s abilities to design de novo protein binders through expert-level orchestration of open-source protein design and structure prediction models. De novo binders are small, computationally designed proteins that attach tightly to a specific target molecule to activate, block, or deliver something to it.","Although this was an encouraging demonstration of AI’s scientific capabilities and an early step towards advancing drug discovery, it took more resources than would be available to the vast majority of protein designers. We allowed Claude to spend up to $10,000 per target on the AI infrastructure platform Modal, roughly equivalent to 2,500 NVIDIA H100 GPU hours.","To make such research more accessible, we began to explore inference optimizations to run these models more efficiently. As an early result of these optimizations, Claude Mythos 5.1：https://www.anthropic.com/claude-fable-and-mythos-5-1 accelerated seven open-source biology models, enabling them to run up to 2.5 times faster.","Here, we present new results showing how an internal, general-purpose research model was able to optimize more than 30 deep learning models trained for a variety of biological tasks, such as structure prediction and protein design, as well as for genomics and protein language models. On average, Claude was able to speed up such tasks roughly 4x while sacrificing a minimal amount of precision, and nearly 2x with identical outputs. Claude also improved the memory utilization of these models, making it possible to predict biomolecular systems of unprecedented sizes. By combining these results with simplifications to our previous agentic protein design approach, we show that Claude can achieve comparable in silico performance to the results we previously reported using two orders of magnitude fewer GPU hours.","Beyond protein design, these specialized biological models are widely used by molecular biologists, including for drug discovery and development. We are open-sourcing the optimized code for all of these models today (here：https://github.com/anthropics/uplifting-biomolecular-modeling) so that the broader community can make use of them. You can find more detail in our technical report (here：https://www-cdn.anthropic.com/d8ca26d0d205708d26c7337cf4cfe7cb52e9b671.pdf).","To further support the community, we are also co-sponsoring a protein design competition with Adaptyv Bio, which has pioneered open protein design competitions：https://proteinbase.com/competitions. We’ve jointly selected five challenging problems at the frontier of today’s capabilities. Together with Adaptyv, and thanks to generous contributions from Modal and Twist Bioscience, we’re committing up to $1 million in Claude credits and $250,000 in Modal compute credits, as well as wet lab validation for over 5,000 designs. Find more information (here：https://proteinbase.com/competitions/anthropic-adaptyv-2026) and (apply here：https://docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDghmEhG_OeVEcj9YGHxLWqxF1kWw/viewform?usp=dialog).","Protein structure prediction is the problem of determining the three-dimensional structure of a protein from its sequence of amino acids alone. Protein design, meanwhile, is the process of creating a protein with a specific structure, function, or set of properties. Together, these computational tools allow scientists to interrogate key biomolecular processes, such as how cancers form, and to create useful molecules, such as drugs that could target these cancers.","Modern structure prediction models, such as AlphaFold3, OpenFold3, and Boltz-2, spend much of their computational runtime and memory on two operations: triangle attention and triangle multiplication, which act on triplets of tokens. These operations make it possible to model the geometry of biomolecular systems, but they are extremely computationally expensive, because they are cubic in both runtime and memory: doubling the size of the system uses 8x more time and memory, while tripling it uses 27x more.","Writing kernels—low-level software translation layers for accelerated computing hardware such as GPUs—is a standard approach for reducing these costs. Given their significance, triangle attention and multiplication have been the subject of dedicated kernel development efforts, first with NVIDIA’s cuEquivariance：https://github.com/nvidia/cuequivariance and more recently with NVIDIA’s BioNeMo Inference Runtime：https://github.com/NVIDIA-BioNeMo/BioNeMo-Inference-Runtime (BioNeMo-IR).","For our own effort to optimize inference for structure prediction models, we worked with Claude to develop FlashPairformer, a set of custom kernels that speed up triangle attention and multiplication. It achieves a new state-of-the-art, outperforming the field standard：https://github.com/nvidia/cuequivariance on average by 2.7-2.9x on triangle attention and 1.7-3.2x on triangle multiplication, depending on the model configuration.","In addition to developing transferable kernels, we pointed Claude at each individual model with the goal of producing more specific optimizations. These included changes like caching redundant recomputed work and simplifying dead branches into their constant outputs. The combination of these improvements accelerated the structure prediction models by 4x, on average, and for each model, we confirmed that Claude’s accelerated versions did not impact performance on the downstream task (such as structure prediction).","It normally takes an experienced team of engineers weeks to produce such optimizations for each model, and the work often does not transfer between models. Claude, supervised by two members of Anthropic’s technical staff who are experienced in biomolecular modeling but who had no prior experience in inference optimization or kernel engineering, carried out the acceleration of more than 30 open-source models across biomolecular structure prediction, protein design, protein language modeling, and genomics in just under four weeks. Our results suggest that frontier AI models will help others in the field build scientific tools with greater speed and ease.","In addition to making these protein structure prediction and design models faster, we also tasked Claude with reducing the memory usage involved in modeling large molecular machines. Much of the work in a cell is done by such systems, including the ribosome that builds proteins, the respiratory complexes that power the cell, and the chaperones that help other proteins fold. Each is built from dozens of components, and its function depends on how those components fit together and interact. Predicting the structures of systems this large has typically required substantial computing resources inaccessible to most molecular biologists, such as inference spread across multiple GPU nodes.","Claude created a low-memory “Big” mode that enables the accurate modeling of systems larger than 10,000 tokens and successful inference on systems larger than 70,000 tokens using just one NVIDIA GPU node—a previously out-of-reach task. Molecular machines folded successfully using Big mode include human mitochondrial complex I, the TRiC chaperone complex, a proteasome, and a bacterial ribosome, each closely matching its experimentally determined structure. To our knowledge, these are among the largest structures ever folded accurately using structure prediction models, with complex I and the 70S ribosome consisting of more than 10,000 tokens each, in comparison to the 40S ribosome predicted accurately by AlphaFold3：https://www.nature.com/articles/s41586-024-07487-w, which consisted of 7,663 tokens.","To test the limits of Claude’s optimizations, we asked Claude to predict structures of a greater size than anything that had previously been achieved. Using a single 8-GPU B300 node, Claude generated predictions of entire viral capsids and protein compartments ranging in size from more than 31,000 to more than 70,000 tokens. These systems are nearly two orders of magnitude larger than the training context of these structure prediction models, and, perhaps unsurprisingly, are not predicted correctly. However, the barrier to inferencing at this scale has been significantly lowered now that it takes just one NVIDIA B300 node, suggesting that with improved tools researchers will soon be able to computationally model an increasingly complex set of biological systems.","In our earlier work on protein design, we provided Claude with an approximately 16,000 word prompt that encouraged it to utilize sub-agents and spend up to $10,000 per target on Modal (roughly 2,500 NVIDIA H100 GPU hours) in a 24 hour span. Here, we gave a single Claude model access to one NVIDIA H200 and 24 hours of wall time, a prompt of about 1,100 words, and a reference sheet for the pre-installed tools, with no sub-agents and no human steering the designs.","We ran three Claude models (Mythos 5.1, Mythos 5, and Opus 5) against 16 targets with the accelerated biomolecular models described in this post. We scored designs by ipSAE：https://www.biorxiv.org/content/10.1101/2025.02.10.637595v2, an in silico score that has been shown to be predictive of binding in the wet lab. Averaged over 16 targets, the median-scoring and highest-scoring designs from all three Claude models evaluated achieve approximately the same ipSAE values as our earlier Mythos 5.1 campaigns despite using about two orders of magnitude fewer GPU hours. We also considered Claude token costs and found that with a combined spend of approximately $150 on GPUs and tokens, we can achieve in silico performance matching the levels of our previous campaigns.","The optimizations described above help us predict and design molecules more efficiently, while unlocking capabilities that would have otherwise been resource-prohibitive. To demonstrate the uplift they provide and the impact of Claude on molecule design more broadly, we’re partnering with Adaptyv Bio to launch a protein design competition：https://proteinbase.com/competitions/anthropic-adaptyv-2026. We’ve selected five problems at the frontier of today’s protein design capabilities, including challenges such as species cross-reactivity, pH-sensitivity, and peptide-MHC specificity, as well as difficult targets such as GPCRs.","With the Adaptyv team, we’ll be experimentally validating over 5,000 designs submitted by the community against these problems. We will be providing up to $1 million in Claude credits and additional funds for experimental validation at Adaptyv for participating researchers, Modal will provide up to $250,000 in compute credits, and Twist Bioscience will provide DNA for the competition. You can find more information, including eligibility criteria (here：https://proteinbase.com/competitions/anthropic-adaptyv-2026) and (apply here：https://docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDghmEhG_OeVEcj9YGHxLWqxF1kWw/viewform?usp=dialog).","We have also begun to provide frontier AI capabilities to life scientists for biology-related work via our Life Sciences Verification Program. We recently enrolled our first group of organizations, and opened up the program in public beta today. You can find more information (here：https://www.anthropic.com/news/life-sciences-verification-program).","The following resources provide further technical depth and more detailed information about the results described above:","Anthropic’s Frontier Red Team developed new evaluations to measure AI capabilities in tactical intelligence targeting and conventional weapons development.","We present an alignment assessment of four incidents in which Claude models gained unauthorized access to real third-party systems.","We are sharing the first complete computer-checked proof of Fermat’s Last Theorem. Claude worked largely autonomously over 11 days to write the proof in the Lean programming language.","Features on AI-assisted discoveries, practical workflows, and field notes across the sciences."],"articleImages":[{"sourceUrl":"https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2F56c4bf34b8eacaa7e1c30648a1d87dd284c9c9b3-2048x1483.jpg&w=3840&q=75","alt":"Chart showing Claude's kernel optimization","afterParagraph":9,"url":"/media/articles/cmu5y1dkt069qroiqnnhxz778/f9161176fc055bfc.webp"},{"sourceUrl":"https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2Fb5ed53a2cb9d20cf369be85a586c1b4d1c171c47-2048x1318.jpg&w=3840&q=75","alt":"Bar chart showing Claude's accelerated optimizations over dozens of structure models","afterParagraph":11,"url":"/media/articles/cmu5y1dkt069qroiqnnhxz778/24f4de59902a7283.webp"},{"sourceUrl":"https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2F55c1085595aeefe9253d497e046b9a007e33d689-2048x1029.jpg&w=3840&q=75","alt":"Bar chart showing how Claude’s optimizations also sped up multiple protein design models spanning hallucination, structure generation, and inverse folding.","afterParagraph":11,"url":"/media/articles/cmu5y1dkt069qroiqnnhxz778/66083460fa2b99c8.webp"},{"sourceUrl":"https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2F7ff86777678b41a6e8852ffbade8737923dd80d3-2048x1323.jpg&w=3840&q=75","alt":"Chart showing the performance of Claude's fast mode","afterParagraph":11,"url":"/media/articles/cmu5y1dkt069qroiqnnhxz778/53d42ca6e947ccc2.webp"},{"sourceUrl":"https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2Ff050c0e37767591e3a7a0f1ea3346515b027f1f6-1908x2048.jpg&w=3840&q=75","alt":"Image of large molecular complexes created by Claude's \"big\" mode","afterParagraph":13,"url":"/media/articles/cmu5y1dkt069qroiqnnhxz778/59d2ab9b7fef5cac.webp"},{"sourceUrl":"https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-cdn.anthropic.com%2Fimages%2F4zrzovbb%2Fwebsite%2F783de4f90edb45cdc1ace96c1fcee7694e7b8c26-2048x680.jpg&w=3840&q=75","alt":"More images of large molecular machines","afterParagraph":14,"url":"/media/articles/cmu5y1dkt069qroiqnnhxz778/8a3e0818559b2566.webp"}],"mediaStatus":"ok","articleBodyZh":["在这篇文章中，我们分享了 Claude 如何使科学家用于预测和设计生物分子的开源模型更快且更节省内存。Claude 在不到四周的时间内优化了 30 多个此类模型，使它们的速度平均提高约 4 倍。它还创建了一个低内存模式，使得在单个 NVIDIA GPU 节点上能够准确预测超过 10,000 个标记（氨基酸、核苷酸以及来自小分子和离子的原子）的生物分子系统。我们正在开源所有优化后的代码，并宣布与 Adaptyv Bio 联合赞助的一场蛋白质设计竞赛，奖励包括价值高达 100 万美元的 Claude 积分以及对 5,000 多个设计进行湿实验验证。最近，我们分享了成果：https://www.anthropic.com/research/Claude-accelerates-protein-design，展示了 Claude 如何通过专家级协调开源蛋白质设计和结构预测模型来设计“de novo”蛋白质结合物。“de novo”结合物是通过计算设计的小型蛋白质，可以紧密附着于特定靶分子以激活、阻断或传递某种物质。","尽管这是一个令人鼓舞的 AI 科学能力展示，也是推动药物发现的早期一步，但这需要的资源远多于大多数蛋白质设计师所能获得的。我们允许 Claude 在 AI 基础设施平台 Modal 上每个靶点使用高达 1 万美元的资源，相当于约 2,500 个 NVIDIA H100 GPU 小时。","为了使这类研究更加可及，我们开始探索推理优化，以更高效地运行这些模型。作为这些优化的早期成果，Claude Mythos 5.1：https://www.anthropic.com/claude-fable-and-mythos-5-1 加速了七个开源生物学模型，使它们的运行速度提高最多 2.5 倍。","在这里，我们展示了新的结果，表明一个通用的内部研究模型能够优化30多个针对各种生物学任务训练的深度学习模型，例如结构预测和蛋白质设计，以及基因组学和蛋白质语言模型。平均而言，Claude 能够将这些任务的速度大约提高4倍，同时仅牺牲极少的精度，并且在输出完全相同的情况下，几乎提高了2倍。Claude 还改善了这些模型的内存利用率，使预测前所未有规模的生物分子系统成为可能。通过将这些结果与我们以前的主动蛋白质设计方法的简化相结合，我们表明 Claude 可以以使用 GPU 小时数减少两个数量级的情况下，实现与我们之前报道的结果相当的体外计算性能。","在蛋白质设计之外，这些专业的生物学模型也被分子生物学家广泛使用，包括药物发现和开发。我们今天将所有这些模型的优化代码开源（这里：https://github.com/anthropics/uplifting-biomolecular-modeling），以便更广泛的社区使用。您可以在我们的技术报告中找到更多细节（这里：https://www-cdn.anthropic.com/d8ca26d0d205708d26c7337cf4cfe7cb52e9b671.pdf）。","为了进一步支持社区，我们还与 Adaptyv Bio 共同赞助了一项蛋白质设计竞赛，该公司在开放蛋白质设计竞赛方面处于领先地位：https://proteinbase.com/competitions。我们联合选择了五个当今能力前沿的挑战性问题。与 Adaptyv 合作，并感谢 Modal 和 Twist Bioscience 的慷慨贡献，我们承诺提供高达100万美元的 Claude 积分和25万美元的 Modal 计算积分，以及超过5000个设计的湿实验验证。更多信息可以参见（这里：https://proteinbase.com/competitions/anthropic-adaptyv-2026）及（申请这里：https://docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDghmEhG_OeVEcj9YGHxLWqxF1kWw/viewform?usp=dialog）。","蛋白质结构预测是仅根据氨基酸序列确定蛋白质三维结构的问题。与此同时，蛋白质设计是创建具有特定结构、功能或属性的蛋白质的过程。这些计算工具共同使科学家能够研究关键生物分子过程，例如癌症如何形成，并创造有用的分子，例如能够靶向这些癌症的药物。","现代结构预测模型，如 AlphaFold3、OpenFold3 和 Boltz-2，将大量计算运行时间和内存消耗用于两个操作：三角注意（triangle attention）和三角乘法（triangle multiplication），这两个操作作用于 token 的三元组。这些操作使得模拟生物分子系统的几何结构成为可能，但它们计算成本极高，因为运行时间和内存都呈三次方增长：系统规模加倍时使用的时间和内存增加 8 倍，三倍时增加 27 倍。","编写内核——针对 GPU 等加速计算硬件的低级软件转换层——是降低这些成本的标准方法。鉴于其重要性，三角注意和三角乘法一直是专门的内核开发目标，最初是 NVIDIA 的 cuEquivariance：https://github.com/nvidia/cuequivariance，最近则是 NVIDIA 的 BioNeMo 推理运行时（BioNeMo-IR）：https://github.com/NVIDIA-BioNeMo/BioNeMo-Inference-Runtime。","在我们优化结构预测模型推理的工作中，我们与 Claude 合作开发了 FlashPairformer，一组加速三角注意和三角乘法的自定义内核。它达到了新的最先进水平，在三角注意上平均超越领域标准：https://github.com/nvidia/cuequivariance 2.7-2.9 倍，三角乘法上根据模型配置超越 1.7-3.2 倍。","除了开发可迁移的内核之外，我们还将Claude指向每个单独的模型，目的是产生更具体的优化。这些优化包括诸如缓存冗余的重复计算工作以及将死分支简化为其常量输出等更改。这些改进的组合平均将结构预测模型的速度提高了4倍，并且对于每个模型，我们确认Claude加速的版本并不影响下游任务（例如结构预测）的性能。","通常，一个经验丰富的工程团队需要数周时间才能为每个模型生成这样的优化，而且这些工作通常无法在不同模型之间迁移。在两名Anthropic技术人员的监督下——他们在生物分子建模方面具有经验，但在推理优化或内核工程方面没有先前经验——Claude在不到四周的时间里完成了对30多个开源模型的加速工作，这些模型涵盖生物分子结构预测、蛋白质设计、蛋白质语言建模和基因组学。我们的结果表明，前沿AI模型将帮助该领域的其他人更快速、更轻松地构建科学工具。","除了加快这些蛋白质结构预测和设计模型的速度之外，我们还让Claude在模拟大型分子机器时减少内存使用。细胞中的大部分工作都是由这些系统完成的，包括构建蛋白质的核糖体、为细胞提供能量的呼吸复合体，以及帮助其他蛋白质折叠的伴侣蛋白。每个系统都是由数十个组件构建的，其功能取决于这些组件如何组合和相互作用。预测如此大型系统的结构通常需要大量计算资源，这些资源对大多数分子生物学家而言是无法获取的，例如分布在多个GPU节点上的推理计算。","Claude 创建了一种低内存“Big”模式，该模式能够精确建模超过 10,000 令牌的大型系统，并能够在仅使用一个 NVIDIA GPU 节点的情况下对超过 70,000 令牌的系统进行成功推理——这是以前难以实现的任务。使用 Big 模式成功折叠的分子机器包括人类线粒体复合物 I、TRiC 分子伴侣复合体、蛋白酶体和细菌核糖体，每个结构都与实验测定的结构高度匹配。据我们所知，这些是迄今为止使用结构预测模型精确折叠的最大结构之一，复合物 I 和 70S 核糖体每个结构都超过 10,000 令牌，而 AlphaFold3 准确预测的 40S 核糖体：https://www.nature.com/articles/s41586-024-07487-w，包含 7,663 令牌。","为了测试 Claude 优化的极限，我们让 Claude 预测比之前任何工作所处理的更大规模的结构。使用单个 8-GPU B300 节点，Claude 生成了整个病毒衣壳和蛋白质隔间的预测，规模从超过 31,000 令牌到超过 70,000 令牌不等。相比这些结构预测模型的训练上下文，这些系统几乎大了两个数量级，并且，或许并不令人意外地，它们的预测结果并不正确。然而，由于现在只需一个 NVIDIA B300 节点就能进行此规模的推理，这一障碍已显著降低，这表明随着工具的改进，研究人员很快将能够计算模拟越来越复杂的生物系统。","在我们早期的蛋白质设计工作中，我们向 Claude 提供了大约 16,000 字的提示词，鼓励它利用子代理，并在 24 小时内每个目标在 Modal 上花费最多 10,000 美元（约 2,500 个 NVIDIA H100 GPU 小时）。在此，我们允许单个 Claude 模型访问一块 NVIDIA H200、24 小时的实际时间，以及大约 1,100 字的提示词和预装工具的参考表，不使用子代理，也没有人类干预设计。","我们使用本文中描述的加速生物分子模型，对16个目标运行了三种Claude模型（Mythos 5.1、Mythos 5 和 Opus 5）。我们通过ipSAE对设计进行评分：https://www.biorxiv.org/content/10.1101/2025.02.10.637595v2，这是一种已被证明能够预测实验室结合性的计算评分。在16个目标上平均，三种Claude模型评估出的中位数评分和最高评分设计与我们早期的Mythos 5.1项目相比，其ipSAE值大致相同，尽管使用的GPU小时数少约两个数量级。我们还考虑了Claude的令牌成本，发现通过约150美元的GPU和令牌总支出，就能实现与我们之前项目水平相当的计算机内性能。","上述优化帮助我们更高效地预测和设计分子，同时解锁了原本因资源限制无法实现的能力。为了展示这些优化带来的提升以及Claude在分子设计中的广泛影响，我们与Adaptyv Bio合作，启动了一项蛋白质设计竞赛：https://proteinbase.com/competitions/anthropic-adaptyv-2026。我们选择了目前蛋白质设计能力前沿的五个问题，包括物种交叉反应性、pH敏感性和肽-MHC特异性等挑战，以及如GPCRs等困难靶点。","与Adaptyv团队合作，我们将对社区提交的超过5000个设计进行实验验证，以应对这些问题。我们将为参与研究人员提供高达100万美元的Claude额度和用于实验验证的额外资金，Modal将提供高达25万美元的计算额度，Twist Bioscience将提供竞赛所需的DNA。您可以在这里找到更多信息，包括资格标准（https://proteinbase.com/competitions/anthropic-adaptyv-2026）以及申请方式（https://docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDghmEhG_OeVEcj9YGHxLWqxF1kWw/viewform?usp=dialog）。","我们还开始通过生命科学验证项目为生命科学家提供前沿AI能力，用于生物学相关工作。我们最近注册了第一批组织，并于今天公开测试阶段开放该项目。更多信息请见：https：//www.anthropic.com/news/life-sciences-verification-program）。","以下资源提供了更深入的技术深度和关于上述结果的详细信息：","Anthropic的前沿红队开发了新的评估，用于衡量AI在战术情报、目标锁定和常规武器开发中的能力。","我们对四起Claude模型未经授权访问真实第三方系统的事件进行了对齐评估。","我们正在分享费马大定理的首个完整的计算机检查证明。Claude在11天内基本自主地用精益编程语言编写了证明。","专题介绍人工智能辅助发现、实用工作流程及跨科学领域的田野笔记。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Anthropic表示，Claude在不到四周内优化了30多个开源生物分子模型，平均提速约4倍；在输出完全一致时接近2倍，并开源全部优化代码。","background":"这些模型覆盖结构预测、蛋白质设计、基因组学和蛋白质语言模型等任务。Anthropic还称，低内存模式可在单张NVIDIA GPU节点上准确预测超过10000个token的生物分子系统。","viewpoint":"Aioga 判断：此次发布的重点不只是单一模型性能提升，还包括将优化代码公开，降低相关生物分子计算的使用门槛；但材料主要呈现Anthropic披露的结果，不能据此推断普遍适用性。","implications":"可能影响：开源优化代码可能为使用相关模型的研究人员提供新的工程参考；但提速与精度表现仍需结合具体模型、任务和运行环境评估，不代表所有生物分子计算都能获得相同收益。","nextStep":"后续观察：应关注开源代码在不同模型和硬件上的复现结果、精度变化及低内存模式的适用边界，也应关注蛋白质设计竞赛及其湿实验验证披露的结果。","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":"2026-09-18T17:17:07.897Z","sourceHash":"63405d2ce5b4ac5c","review":{"approved":true,"groundedness":94,"clarity":89,"duplicationRisk":32,"blockingIssues":[],"notes":["“开源全部优化代码”与来源中“为所有这些模型开源优化代码”一致，可考虑改为“为这些模型开源全部优化代码”以避免范围歧义。","“降低相关生物分子计算的使用门槛”属于基于开源举措的合理判断，已明确标注为观点或可能影响，无需阻断。","可在后续报道中补充具体模型、硬件和任务条件，以进一步限定提速与精度结论的适用范围。"]},"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","Anthropic：Research（发表成果 · 网页）"],"translations":{"zh-CN":{"title":"Anthropic 用 Claude 优化 30 多个开源生物分子模型，平均提速约 4 倍并开源全部代码","summary":"Anthropic 发布研究，让 Claude 在不到四周内优化了 30 多个开源生物分子模型，平均提速约 4 倍，输出完全一致时约 2 倍。","category":"行业动态","source":"Anthropic","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic 用 Claude 优化 30 多个开源生物分子模型，平均提速约 4 倍并开源全部代码 - Aioga AI资讯","description":"Anthropic 发布研究，让 Claude 在不到四周内优化了 30 多个开源生物分子模型，平均提速约 4 倍，输出完全一致时约 2 倍。","url":"https://www.aioga.com/news/cmu5y1dkt069qroiqnnhxz778/","articleBody":["在这篇文章中，我们分享了 Claude 如何使科学家用于预测和设计生物分子的开源模型更快且更节省内存。Claude 在不到四周的时间内优化了 30 多个此类模型，使它们的速度平均提高约 4 倍。它还创建了一个低内存模式，使得在单个 NVIDIA GPU 节点上能够准确预测超过 10,000 个标记（氨基酸、核苷酸以及来自小分子和离子的原子）的生物分子系统。我们正在开源所有优化后的代码，并宣布与 Adaptyv Bio 联合赞助的一场蛋白质设计竞赛，奖励包括价值高达 100 万美元的 Claude 积分以及对 5,000 多个设计进行湿实验验证。最近，我们分享了成果：https://www.anthropic.com/research/Claude-accelerates-protein-design，展示了 Claude 如何通过专家级协调开源蛋白质设计和结构预测模型来设计“de novo”蛋白质结合物。“de novo”结合物是通过计算设计的小型蛋白质，可以紧密附着于特定靶分子以激活、阻断或传递某种物质。","尽管这是一个令人鼓舞的 AI 科学能力展示，也是推动药物发现的早期一步，但这需要的资源远多于大多数蛋白质设计师所能获得的。我们允许 Claude 在 AI 基础设施平台 Modal 上每个靶点使用高达 1 万美元的资源，相当于约 2,500 个 NVIDIA H100 GPU 小时。","为了使这类研究更加可及，我们开始探索推理优化，以更高效地运行这些模型。作为这些优化的早期成果，Claude Mythos 5.1：https://www.anthropic.com/claude-fable-and-mythos-5-1 加速了七个开源生物学模型，使它们的运行速度提高最多 2.5 倍。","在这里，我们展示了新的结果，表明一个通用的内部研究模型能够优化30多个针对各种生物学任务训练的深度学习模型，例如结构预测和蛋白质设计，以及基因组学和蛋白质语言模型。平均而言，Claude 能够将这些任务的速度大约提高4倍，同时仅牺牲极少的精度，并且在输出完全相同的情况下，几乎提高了2倍。Claude 还改善了这些模型的内存利用率，使预测前所未有规模的生物分子系统成为可能。通过将这些结果与我们以前的主动蛋白质设计方法的简化相结合，我们表明 Claude 可以以使用 GPU 小时数减少两个数量级的情况下，实现与我们之前报道的结果相当的体外计算性能。","在蛋白质设计之外，这些专业的生物学模型也被分子生物学家广泛使用，包括药物发现和开发。我们今天将所有这些模型的优化代码开源（这里：https://github.com/anthropics/uplifting-biomolecular-modeling），以便更广泛的社区使用。您可以在我们的技术报告中找到更多细节（这里：https://www-cdn.anthropic.com/d8ca26d0d205708d26c7337cf4cfe7cb52e9b671.pdf）。","为了进一步支持社区，我们还与 Adaptyv Bio 共同赞助了一项蛋白质设计竞赛，该公司在开放蛋白质设计竞赛方面处于领先地位：https://proteinbase.com/competitions。我们联合选择了五个当今能力前沿的挑战性问题。与 Adaptyv 合作，并感谢 Modal 和 Twist Bioscience 的慷慨贡献，我们承诺提供高达100万美元的 Claude 积分和25万美元的 Modal 计算积分，以及超过5000个设计的湿实验验证。更多信息可以参见（这里：https://proteinbase.com/competitions/anthropic-adaptyv-2026）及（申请这里：https://docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDghmEhG_OeVEcj9YGHxLWqxF1kWw/viewform?usp=dialog）。","蛋白质结构预测是仅根据氨基酸序列确定蛋白质三维结构的问题。与此同时，蛋白质设计是创建具有特定结构、功能或属性的蛋白质的过程。这些计算工具共同使科学家能够研究关键生物分子过程，例如癌症如何形成，并创造有用的分子，例如能够靶向这些癌症的药物。","现代结构预测模型，如 AlphaFold3、OpenFold3 和 Boltz-2，将大量计算运行时间和内存消耗用于两个操作：三角注意（triangle attention）和三角乘法（triangle multiplication），这两个操作作用于 token 的三元组。这些操作使得模拟生物分子系统的几何结构成为可能，但它们计算成本极高，因为运行时间和内存都呈三次方增长：系统规模加倍时使用的时间和内存增加 8 倍，三倍时增加 27 倍。","编写内核——针对 GPU 等加速计算硬件的低级软件转换层——是降低这些成本的标准方法。鉴于其重要性，三角注意和三角乘法一直是专门的内核开发目标，最初是 NVIDIA 的 cuEquivariance：https://github.com/nvidia/cuequivariance，最近则是 NVIDIA 的 BioNeMo 推理运行时（BioNeMo-IR）：https://github.com/NVIDIA-BioNeMo/BioNeMo-Inference-Runtime。","在我们优化结构预测模型推理的工作中，我们与 Claude 合作开发了 FlashPairformer，一组加速三角注意和三角乘法的自定义内核。它达到了新的最先进水平，在三角注意上平均超越领域标准：https://github.com/nvidia/cuequivariance 2.7-2.9 倍，三角乘法上根据模型配置超越 1.7-3.2 倍。","除了开发可迁移的内核之外，我们还将Claude指向每个单独的模型，目的是产生更具体的优化。这些优化包括诸如缓存冗余的重复计算工作以及将死分支简化为其常量输出等更改。这些改进的组合平均将结构预测模型的速度提高了4倍，并且对于每个模型，我们确认Claude加速的版本并不影响下游任务（例如结构预测）的性能。","通常，一个经验丰富的工程团队需要数周时间才能为每个模型生成这样的优化，而且这些工作通常无法在不同模型之间迁移。在两名Anthropic技术人员的监督下——他们在生物分子建模方面具有经验，但在推理优化或内核工程方面没有先前经验——Claude在不到四周的时间里完成了对30多个开源模型的加速工作，这些模型涵盖生物分子结构预测、蛋白质设计、蛋白质语言建模和基因组学。我们的结果表明，前沿AI模型将帮助该领域的其他人更快速、更轻松地构建科学工具。","除了加快这些蛋白质结构预测和设计模型的速度之外，我们还让Claude在模拟大型分子机器时减少内存使用。细胞中的大部分工作都是由这些系统完成的，包括构建蛋白质的核糖体、为细胞提供能量的呼吸复合体，以及帮助其他蛋白质折叠的伴侣蛋白。每个系统都是由数十个组件构建的，其功能取决于这些组件如何组合和相互作用。预测如此大型系统的结构通常需要大量计算资源，这些资源对大多数分子生物学家而言是无法获取的，例如分布在多个GPU节点上的推理计算。","Claude 创建了一种低内存“Big”模式，该模式能够精确建模超过 10,000 令牌的大型系统，并能够在仅使用一个 NVIDIA GPU 节点的情况下对超过 70,000 令牌的系统进行成功推理——这是以前难以实现的任务。使用 Big 模式成功折叠的分子机器包括人类线粒体复合物 I、TRiC 分子伴侣复合体、蛋白酶体和细菌核糖体，每个结构都与实验测定的结构高度匹配。据我们所知，这些是迄今为止使用结构预测模型精确折叠的最大结构之一，复合物 I 和 70S 核糖体每个结构都超过 10,000 令牌，而 AlphaFold3 准确预测的 40S 核糖体：https://www.nature.com/articles/s41586-024-07487-w，包含 7,663 令牌。","为了测试 Claude 优化的极限，我们让 Claude 预测比之前任何工作所处理的更大规模的结构。使用单个 8-GPU B300 节点，Claude 生成了整个病毒衣壳和蛋白质隔间的预测，规模从超过 31,000 令牌到超过 70,000 令牌不等。相比这些结构预测模型的训练上下文，这些系统几乎大了两个数量级，并且，或许并不令人意外地，它们的预测结果并不正确。然而，由于现在只需一个 NVIDIA B300 节点就能进行此规模的推理，这一障碍已显著降低，这表明随着工具的改进，研究人员很快将能够计算模拟越来越复杂的生物系统。","在我们早期的蛋白质设计工作中，我们向 Claude 提供了大约 16,000 字的提示词，鼓励它利用子代理，并在 24 小时内每个目标在 Modal 上花费最多 10,000 美元（约 2,500 个 NVIDIA H100 GPU 小时）。在此，我们允许单个 Claude 模型访问一块 NVIDIA H200、24 小时的实际时间，以及大约 1,100 字的提示词和预装工具的参考表，不使用子代理，也没有人类干预设计。","我们使用本文中描述的加速生物分子模型，对16个目标运行了三种Claude模型（Mythos 5.1、Mythos 5 和 Opus 5）。我们通过ipSAE对设计进行评分：https://www.biorxiv.org/content/10.1101/2025.02.10.637595v2，这是一种已被证明能够预测实验室结合性的计算评分。在16个目标上平均，三种Claude模型评估出的中位数评分和最高评分设计与我们早期的Mythos 5.1项目相比，其ipSAE值大致相同，尽管使用的GPU小时数少约两个数量级。我们还考虑了Claude的令牌成本，发现通过约150美元的GPU和令牌总支出，就能实现与我们之前项目水平相当的计算机内性能。","上述优化帮助我们更高效地预测和设计分子，同时解锁了原本因资源限制无法实现的能力。为了展示这些优化带来的提升以及Claude在分子设计中的广泛影响，我们与Adaptyv Bio合作，启动了一项蛋白质设计竞赛：https://proteinbase.com/competitions/anthropic-adaptyv-2026。我们选择了目前蛋白质设计能力前沿的五个问题，包括物种交叉反应性、pH敏感性和肽-MHC特异性等挑战，以及如GPCRs等困难靶点。","与Adaptyv团队合作，我们将对社区提交的超过5000个设计进行实验验证，以应对这些问题。我们将为参与研究人员提供高达100万美元的Claude额度和用于实验验证的额外资金，Modal将提供高达25万美元的计算额度，Twist Bioscience将提供竞赛所需的DNA。您可以在这里找到更多信息，包括资格标准（https://proteinbase.com/competitions/anthropic-adaptyv-2026）以及申请方式（https://docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDghmEhG_OeVEcj9YGHxLWqxF1kWw/viewform?usp=dialog）。","我们还开始通过生命科学验证项目为生命科学家提供前沿AI能力，用于生物学相关工作。我们最近注册了第一批组织，并于今天公开测试阶段开放该项目。更多信息请见：https：//www.anthropic.com/news/life-sciences-verification-program）。","以下资源提供了更深入的技术深度和关于上述结果的详细信息：","Anthropic的前沿红队开发了新的评估，用于衡量AI在战术情报、目标锁定和常规武器开发中的能力。","我们对四起Claude模型未经授权访问真实第三方系统的事件进行了对齐评估。","我们正在分享费马大定理的首个完整的计算机检查证明。Claude在11天内基本自主地用精益编程语言编写了证明。","专题介绍人工智能辅助发现、实用工作流程及跨科学领域的田野笔记。"]},"en":{"title":"Anthropic optimized more than 30 open-source biomolecular models using Claude, achieving an average speed increase of about 4 times and open-sourced all the code.","summary":"Anthropic released a study showing that Claude optimized more than 30 open-source biomolecular models in less than four weeks, with an average speed-up of about 4 times, and about 2 times when the output was completely consistent.","category":"Industry","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic optimized more than 30 open-source biomolecular models using Claude, achieving an average speed increase of about 4 times and open-sourced all the code. - Aioga AI News","description":"Anthropic released a study showing that Claude optimized more than 30 open-source biomolecular models in less than four weeks, with an average speed-up of about 4 times, and about...","url":"https://www.aioga.com/en/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:13:15.412Z"},"ja":{"title":"Anthropic は Claude を用いて30以上のオープンソースの生体分子モデルを最適化し、平均で約4倍高速化し、全てのコードをオープンソース化しました","summary":"Anthropic は研究を発表し、Claude がわずか4週間足らずで30以上のオープンソース生体分子モデルを最適化し、平均で約4倍の速度向上、出力が完全に一致する場合は約2倍となった。","category":"業界動向","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic は Claude を用いて30以上のオープンソースの生体分子モデルを最適化し、平均で約4倍高速化し、全てのコードをオープンソース化しました - Aioga AIニュース","description":"Anthropic は研究を発表し、Claude がわずか4週間足らずで30以上のオープンソース生体分子モデルを最適化し、平均で約4倍の速度向上、出力が完全に一致する場合は約2倍となった。","url":"https://www.aioga.com/ja/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:13:19.330Z"},"ko":{"title":"Anthropic는 Claude를 사용하여 30개 이상의 오픈소스 생체 분자 모델을 최적화했으며, 평균 속도를 약 4배 향상시키고 모든 코드를 오픈소스로 제공했습니다.","summary":"Anthropic는 연구를 발표하여 Claude가 4주도 안 되게 30개 이상의 오픈소스 생체 분자 모델을 최적화했으며, 평균 속도는 약 4배, 출력이 완전히 일치할 때는 약 2배 빨라졌습니다.","category":"업계 동향","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic는 Claude를 사용하여 30개 이상의 오픈소스 생체 분자 모델을 최적화했으며, 평균 속도를 약 4배 향상시키고 모든 코드를 오픈소스로 제공했습니다. - Aioga AI 뉴스","description":"Anthropic는 연구를 발표하여 Claude가 4주도 안 되게 30개 이상의 오픈소스 생체 분자 모델을 최적화했으며, 평균 속도는 약 4배, 출력이 완전히 일치할 때는 약 2배 빨라졌습니다.","url":"https://www.aioga.com/ko/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:14:21.789Z"},"es":{"title":"Anthropic usó Claude para optimizar más de 30 modelos de biomoléculas de código abierto, aumentando la velocidad en aproximadamente 4 veces y liberando todo el código.","summary":"Anthropic publicó un estudio en el que Claude optimizó en menos de cuatro semanas más de 30 modelos de biomoléculas de código abierto, con un aumento promedio de velocidad de aproximadamente 4 veces, y aproximadamente 2 veces cuando la salida era completamente consistente.","category":"Industria","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic usó Claude para optimizar más de 30 modelos de biomoléculas de código abierto, aumentando la velocidad en aproximadamente 4 veces y liberando todo el código. - Aioga Noticias de IA","description":"Anthropic publicó un estudio en el que Claude optimizó en menos de cuatro semanas más de 30 modelos de biomoléculas de código abierto, con un aumento promedio de velocidad de aprox...","url":"https://www.aioga.com/es/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:14:25.650Z"},"fr":{"title":"Anthropic a utilisé Claude pour optimiser plus de 30 modèles moléculaires biologiques open source, augmentant la vitesse en moyenne d'environ 4 fois et publiant tout le code.","summary":"Anthropic a publié une recherche montrant que Claude a optimisé plus de 30 modèles de biomolécules open source en moins de quatre semaines, avec une accélération moyenne d'environ 4 fois, et environ 2 fois lorsque la sortie est entièrement identique.","category":"Industrie","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic a utilisé Claude pour optimiser plus de 30 modèles moléculaires biologiques open source, augmentant la vitesse en moyenne d'environ 4 fois et publiant tout le code. - Aioga Actualités IA","description":"Anthropic a publié une recherche montrant que Claude a optimisé plus de 30 modèles de biomolécules open source en moins de quatre semaines, avec une accélération moyenne d'environ...","url":"https://www.aioga.com/fr/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:15:26.458Z"},"de":{"title":"Anthropic hat über 30 Open-Source-Biomolekülmodelle mit Claude optimiert, die durchschnittliche Geschwindigkeit um etwa das Vierfache erhöht und den gesamten Code Open Source gemacht.","summary":"Anthropic veröffentlichte eine Studie, in der Claude in weniger als vier Wochen über 30 Open-Source-Biomolekülmodelle optimierte, mit einer durchschnittlichen Beschleunigung von etwa dem Vierfachen und etwa dem Doppelten bei vollständig identischem Output.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic hat über 30 Open-Source-Biomolekülmodelle mit Claude optimiert, die durchschnittliche Geschwindigkeit um etwa das Vierfache erhöht und den gesamten Code Open Source gemacht. - Aioga KI-News","description":"Anthropic veröffentlichte eine Studie, in der Claude in weniger als vier Wochen über 30 Open-Source-Biomolekülmodelle optimierte, mit einer durchschnittlichen Beschleunigung von et...","url":"https://www.aioga.com/de/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:15:25.704Z"},"pt-BR":{"title":"A Anthropic usou o Claude para otimizar mais de 30 modelos moleculares biológicos de código aberto, acelerando em média cerca de 4 vezes e tornando todo o código open source.","summary":"A Anthropic publicou uma pesquisa que permitiu que o Claude otimizasse mais de 30 modelos moleculares biológicos open-source em menos de quatro semanas, com um aumento médio de velocidade de cerca de 4 vezes e cerca de 2 vezes quando a saída era totalmente consistente.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"A Anthropic usou o Claude para otimizar mais de 30 modelos moleculares biológicos de código aberto, acelerando em média cerca de 4 vezes e tornando todo o código open source. - Aioga Notícias de IA","description":"A Anthropic publicou uma pesquisa que permitiu que o Claude otimizasse mais de 30 modelos moleculares biológicos open-source em menos de quatro semanas, com um aumento médio de vel...","url":"https://www.aioga.com/pt-BR/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:16:30.222Z"},"ru":{"title":"Anthropic оптимизировала более 30 открытых молекулярных моделей с помощью Claude, в среднем ускорив их примерно в 4 раза, и открыла весь код","summary":"Anthropic опубликовала исследование, в котором Claude за менее чем четыре недели оптимизировал более 30 открытых моделей биомолекул, ускорив их в среднем примерно в 4 раза, а при полностью совпадающем выводе примерно в 2 раза.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic оптимизировала более 30 открытых молекулярных моделей с помощью Claude, в среднем ускорив их примерно в 4 раза, и открыла весь код - Aioga Новости ИИ","description":"Anthropic опубликовала исследование, в котором Claude за менее чем четыре недели оптимизировал более 30 открытых моделей биомолекул, ускорив их в среднем примерно в 4 раза, а при п...","url":"https://www.aioga.com/ru/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:16:32.547Z"},"ar":{"title":"شركة Anthropic استخدمت Claude لتحسين أكثر من 30 نموذج جزيئي حيوي مفتوح المصدر، بزيادة سرعة متوسطة حوالي 4 أضعاف وفتحت جميع الأكواد المصدرية","summary":"أصدرت Anthropic بحثًا، سمح لـ Claude بتحسين أكثر من 30 نموذجًا مفتوح المصدر للجزيئات الحيوية في أقل من أربعة أسابيع، بمعدل تسريع حوالي 4 مرات في المتوسط، وحوالي مرتين عند التوافق الكامل في المخرجات.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"شركة Anthropic استخدمت Claude لتحسين أكثر من 30 نموذج جزيئي حيوي مفتوح المصدر، بزيادة سرعة متوسطة حوالي 4 أضعاف وفتحت جميع الأكواد المصدرية - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت Anthropic بحثًا، سمح لـ Claude بتحسين أكثر من 30 نموذجًا مفتوح المصدر للجزيئات الحيوية في أقل من أربعة أسابيع، بمعدل تسريع حوالي 4 مرات في المتوسط، وحوالي مرتين عند التوافق ا...","url":"https://www.aioga.com/ar/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:17:32.497Z"},"hi":{"title":"Anthropic ने Claude का उपयोग करके 30 से अधिक ओपन-सोर्स जैव अणु मॉडलों को अनुकूलित किया, औसत गति लगभग 4 गुना बढ़ी और सभी कोड को ओपन-सोर्स किया","summary":"Anthropic ने एक अध्ययन प्रकाशित किया, जिससे Claude ने चार सप्ताह से कम समय में 30 से अधिक ओपन-सोर्स बायोमोलिक्यूल मॉडल्स का अनुकूलन किया, औसतन लगभग 4 गुना तेजी से, और आउटपुट पूरी तरह समान होने पर लगभग 2 गुना।","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic ने Claude का उपयोग करके 30 से अधिक ओपन-सोर्स जैव अणु मॉडलों को अनुकूलित किया, औसत गति लगभग 4 गुना बढ़ी और सभी कोड को ओपन-सोर्स किया - Aioga AI समाचार","description":"Anthropic ने एक अध्ययन प्रकाशित किया, जिससे Claude ने चार सप्ताह से कम समय में 30 से अधिक ओपन-सोर्स बायोमोलिक्यूल मॉडल्स का अनुकूलन किया, औसतन लगभग 4 गुना तेजी से, और आउटपुट पूरी त...","url":"https://www.aioga.com/hi/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:17:33.834Z"},"it":{"title":"Anthropic ha ottimizzato oltre 30 modelli molecolari biologici open source con Claude, aumentando in media la velocità di circa 4 volte e ha reso open source tutto il codice","summary":"Anthropic ha pubblicato una ricerca, facendo sì che Claude ottimizzasse più di 30 modelli molecolari biologici open source in meno di quattro settimane, con una velocità media circa 4 volte superiore e circa 2 volte quando l'output era completamente coerente.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic ha ottimizzato oltre 30 modelli molecolari biologici open source con Claude, aumentando in media la velocità di circa 4 volte e ha reso open source tutto il codice - Aioga Notizie IA","description":"Anthropic ha pubblicato una ricerca, facendo sì che Claude ottimizzasse più di 30 modelli molecolari biologici open source in meno di quattro settimane, con una velocità media circ...","url":"https://www.aioga.com/it/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:18:31.180Z"},"nl":{"title":"Anthropic heeft met Claude meer dan 30 open-source biomoleculaire modellen geoptimaliseerd, wat een gemiddelde snelheidsverbetering van ongeveer 4 keer oplevert en alle code is open-source gemaakt","summary":"Anthropic publiceert onderzoek waarin Claude in minder dan vier weken meer dan 30 open source biomoleculaire modellen optimaliseerde, met een gemiddelde snelheidsverbetering van ongeveer 4 keer, en ongeveer 2 keer wanneer de uitvoer volledig identiek is.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic heeft met Claude meer dan 30 open-source biomoleculaire modellen geoptimaliseerd, wat een gemiddelde snelheidsverbetering van ongeveer 4 keer oplevert en alle code is open-source gemaakt - Aioga AI-nieuws","description":"Anthropic publiceert onderzoek waarin Claude in minder dan vier weken meer dan 30 open source biomoleculaire modellen optimaliseerde, met een gemiddelde snelheidsverbetering van on...","url":"https://www.aioga.com/nl/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:18:31.433Z"},"tr":{"title":"Anthropic, Claude'u kullanarak 30'dan fazla açık kaynak biyomoleküler modeli optimize etti, ortalama hız artışı yaklaşık 4 kat sağladı ve tüm kodları açık kaynak yaptı","summary":"Anthropic, Claude'un dört haftadan kısa sürede 30'dan fazla açık kaynak biyomoleküler modeli optimize etmesini sağlayan bir araştırma yayınladı; ortalama hız artışı yaklaşık 4 kat, çıktı tamamen tutarlı olduğunda ise yaklaşık 2 kat oldu.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic, Claude'u kullanarak 30'dan fazla açık kaynak biyomoleküler modeli optimize etti, ortalama hız artışı yaklaşık 4 kat sağladı ve tüm kodları açık kaynak yaptı - Aioga AI Haberleri","description":"Anthropic, Claude'un dört haftadan kısa sürede 30'dan fazla açık kaynak biyomoleküler modeli optimize etmesini sağlayan bir araştırma yayınladı; ortalama hız artışı yaklaşık 4 kat,...","url":"https://www.aioga.com/tr/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:19:39.103Z"},"vi":{"title":"Anthropic đã sử dụng Claude để tối ưu hóa hơn 30 mô hình phân tử sinh học mã nguồn mở, tăng tốc trung bình khoảng 4 lần và công khai toàn bộ mã nguồn","summary":"Anthropic đã công bố nghiên cứu, giúp Claude tối ưu hóa hơn 30 mô hình phân tử sinh học mã nguồn mở trong chưa đầy bốn tuần, tăng tốc trung bình khoảng 4 lần, và khi đầu ra hoàn toàn nhất quán thì khoảng 2 lần.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic đã sử dụng Claude để tối ưu hóa hơn 30 mô hình phân tử sinh học mã nguồn mở, tăng tốc trung bình khoảng 4 lần và công khai toàn bộ mã nguồn - Tin tức AI Aioga","description":"Anthropic đã công bố nghiên cứu, giúp Claude tối ưu hóa hơn 30 mô hình phân tử sinh học mã nguồn mở trong chưa đầy bốn tuần, tăng tốc trung bình khoảng 4 lần, và khi đầu ra hoàn to...","url":"https://www.aioga.com/vi/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:19:46.586Z"},"id":{"title":"Anthropic menggunakan Claude untuk mengoptimalkan lebih dari 30 model biomolekuler open source, dengan rata-rata percepatan sekitar 4 kali dan merilis seluruh kode secara open source","summary":"Anthropic merilis penelitian, membuat Claude mengoptimalkan lebih dari 30 model molekul biologi sumber terbuka dalam waktu kurang dari empat minggu, dengan peningkatan kecepatan rata-rata sekitar 4 kali, dan ketika output sepenuhnya konsisten sekitar 2 kali lipat.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic menggunakan Claude untuk mengoptimalkan lebih dari 30 model biomolekuler open source, dengan rata-rata percepatan sekitar 4 kali dan merilis seluruh kode secara open source - Berita AI Aioga","description":"Anthropic merilis penelitian, membuat Claude mengoptimalkan lebih dari 30 model molekul biologi sumber terbuka dalam waktu kurang dari empat minggu, dengan peningkatan kecepatan ra...","url":"https://www.aioga.com/id/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:20:46.918Z"},"th":{"title":"Anthropic ใช้ Claude ปรับปรุงโมเดลโมเลกุลทางชีวภาพโอเพนซอร์สมากกว่า 30 โมเดล เร็วขึ้นโดยเฉลี่ยประมาณ 4 เท่าและเปิดโค้ดทั้งหมด","summary":"Anthropic ได้เผยแพร่การวิจัย ทำให้ Claude ปรับปรุงโมเดลชีวโมเลกุลโอเพนซอร์สมากกว่า 30 โมเดลในเวลาไม่ถึงสี่สัปดาห์ โดยความเร็วเฉลี่ยเพิ่มขึ้นประมาณ 4 เท่า และเมื่อผลลัพธ์ตรงกันทั้งหมดจะเพิ่มขึ้นประมาณ 2 เท่า","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic ใช้ Claude ปรับปรุงโมเดลโมเลกุลทางชีวภาพโอเพนซอร์สมากกว่า 30 โมเดล เร็วขึ้นโดยเฉลี่ยประมาณ 4 เท่าและเปิดโค้ดทั้งหมด - ข่าว AI Aioga","description":"Anthropic ได้เผยแพร่การวิจัย ทำให้ Claude ปรับปรุงโมเดลชีวโมเลกุลโอเพนซอร์สมากกว่า 30 โมเดลในเวลาไม่ถึงสี่สัปดาห์ โดยความเร็วเฉลี่ยเพิ่มขึ้นประมาณ 4 เท่า และเมื่อผลลัพธ์ตรงกันทั้งห...","url":"https://www.aioga.com/th/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:20:53.177Z"},"pl":{"title":"Anthropic wykorzystał Claude do optymalizacji ponad 30 otwartoźródłowych modeli cząsteczek biologicznych, średnio przyspieszając je około 4 razy i udostępniając cały kod źródłowy","summary":"Anthropic opublikował badania, które pozwoliły Claudowi w mniej niż cztery tygodnie zoptymalizować ponad 30 otwartych modeli biomolekularnych, przy średnim przyspieszeniu około 4 razy, a przy całkowitej zgodności wyników około 2 razy.","category":"行业动态","source":"Anthropic：Research（发表成果 · 网页）","aggregationSource":"Anthropic：Research（发表成果 · 网页）","pageTitle":"Anthropic wykorzystał Claude do optymalizacji ponad 30 otwartoźródłowych modeli cząsteczek biologicznych, średnio przyspieszając je około 4 razy i udostępniając cały kod źródłowy - Aioga Wiadomości AI","description":"Anthropic opublikował badania, które pozwoliły Claudowi w mniej niż cztery tygodnie zoptymalizować ponad 30 otwartych modeli biomolekularnych, przy średnim przyspieszeniu około 4 r...","url":"https://www.aioga.com/pl/news/cmu5y1dkt069qroiqnnhxz778/","contentTranslated":true,"sourceHash":"3e2df66fbe535d62","translatedAt":"2026-09-17T20:22:02.289Z"}},"evidenceTier":"verified-news","reviewStatus":"editorial-selected","indexable":true,"editorialCover":""}}