Anthropic raises Claude Code usage limits, credits new deal with SpaceX
Anthropic raises Claude Code usage limits, credits new deal with SpaceX
Anthropic 提高 Claude Code 使用限额，归功于与 SpaceX 的新协议
SAN FRANCISCO—At its Code with Claude developer conference on Wednesday, Anthropic announced a deal with SpaceX to utilize the entire compute capacity of the latter’s data center in Memphis, Tennessee. On stage at the conference, CEO Dario Amodei said the deal was intended to increase usage limits for Anthropic’s Pro and Max plan subscribers.
旧金山——在周三举行的“Code with Claude”开发者大会上，Anthropic 宣布与 SpaceX 达成协议，将利用后者位于田纳西州孟菲斯数据中心的全部计算能力。首席执行官 Dario Amodei 在大会舞台上表示，此举旨在提高 Anthropic Pro 和 Max 订阅用户的使用限额。
The announcement was accompanied by an increase in those usage limits; Anthropic doubled Claude Code’s five-hour window limits for Pro and Max subscribers, removed the peak-hours limit reduction on Claude Code for those same accounts, and raised API limits for its Opus model.
随着这一公告的发布，相关使用限额也随之提升：Anthropic 将 Pro 和 Max 订阅用户的 Claude Code 五小时窗口期限额翻了一番，取消了这些账户在高峰时段的限额缩减，并提高了 Opus 模型的 API 调用限额。
Anthropic claims the deal gives the company access to more than 300 megawatts of new compute capacity. For its part, SpaceX focused its announcement on the capability of the Colossus 1 supercomputer that’s at the center of the deal. “Colossus 1 features over 220,000 NVIDIA GPUs, including dense deployments of H100, H200, and next-generation GB200 accelerators,” SpaceX 写道。
Anthropic 声称，该协议使其获得了超过 300 兆瓦的新增计算能力。SpaceX 则在公告中重点介绍了作为此次合作核心的 Colossus 1 超级计算机的性能。SpaceX 写道：“Colossus 1 配备了超过 22 万个 NVIDIA GPU，包括高密度部署的 H100、H200 以及下一代 GB200 加速器。”
Additionally, Anthropic “expressed interest” in working with SpaceX to build up “multiple gigawatts” of orbital compute capacity, tying into a recent (but unproven) focus on exploring orbital data centers as an answer to the problem that “compute required to train and operate the next generation of these systems is outpacing what terrestrial power, land, and cooling can deliver on the timelines that matter.”
此外，Anthropic 还“表达了兴趣”，希望与 SpaceX 合作建立“数吉瓦（gigawatts）”级别的轨道计算能力。这与近期（尽管尚未得到证实）探索轨道数据中心的趋势相呼应，旨在解决“下一代系统训练和运行所需的计算能力，正超过地面电力、土地和冷却设施在关键时间节点所能提供的极限”这一问题。
The deal might be surprising to those who have followed Musk’s recent public comments—he was, until now, critical of Anthropic. For example, in February, he declared on X that “Anthropic hates Western Civilization,” while sharing a false tweet from Trump administration official Emil Michael about Anthropic’s practices with its constitution for Claude.
对于那些关注马斯克近期公开言论的人来说，这项协议可能会令人感到意外——因为直到现在，他一直对 Anthropic 持批评态度。例如，今年二月，他在 X 上宣称“Anthropic 仇视西方文明”，并转发了特朗普政府官员 Emil Michael 关于 Anthropic 在 Claude 宪法实践方面的不实推文。
The tune changed with the deal—or in the lead-up to it, as Musk tells it. “I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed,” Musk tweeted on Wednesday. “No one set off my evil detector.”
随着协议的达成，马斯克的态度发生了转变，或者用他自己的话说，是在协议达成前的准备阶段转变的。马斯克周三在推文中写道：“上周我花了很多时间与 Anthropic 团队的高层成员交流，了解他们如何确保 Claude 对人类有益，我印象深刻。没有人触发我的‘邪恶探测器’。”
Exploding demand amid constrained compute supply
计算资源受限背景下的需求激增
Anthropic has seen a significant increase in demand over the past few months for Claude Code and other products related to its models. The increase has been driven partly by users who are moving away from OpenAI after controversy over its agreements with the United States military, increasing adoption of Claude Code in professional software development organizations, and a user behavior (and product) shift away from single-agent, chat-based tasks to more demanding multi-agent workflows.
过去几个月里，Anthropic 的 Claude Code 及其他相关模型产品的需求显著增长。这种增长的部分原因是，在 OpenAI 与美国军方达成协议引发争议后，部分用户开始转向其他平台；同时，Claude Code 在专业软件开发组织中的采用率不断提高，且用户行为（及产品形态）正从单一代理的聊天任务转向要求更高的多代理工作流。
The company has made controversial moves lately to address demand outpacing available compute capacity amid outages and other problems. That included introducing new usage limits during peak hours, and even a short-lived and very limited trial, apparently testing the notion of removing Claude Code from the $20/month Pro plan.
为了应对需求超过现有计算能力导致的宕机及其他问题，该公司近期采取了一些颇具争议的举措。其中包括在高峰时段引入新的使用限额，甚至进行了一次短暂且范围极小的测试，似乎是在评估将 Claude Code 从每月 20 美元的 Pro 套餐中移除的可行性。
Vocal frustration with usage limits has been a staple of Hacker News, Reddit, X, and other platforms where software developers congregate. Many developers are using these models and tools, and they’re frustrated that they can’t use them more. Last month, Anthropic reportedly signed massive deals with Microsoft, Google, Amazon, Nvidia, and more to scale up its access to compute infrastructure. The company credits those, along with this SpaceX deal, for its ability to raise limits, though gains from some of them will take time to materialize.
在 Hacker News、Reddit、X 以及其他软件开发者聚集的平台上，对使用限额的抱怨声不绝于耳。许多开发者都在使用这些模型和工具，但因无法获得更多使用权限而感到沮丧。据报道，上个月 Anthropic 与微软、谷歌、亚马逊、英伟达等公司签署了大规模协议，以扩大其计算基础设施的接入能力。该公司将此次限额提升归功于这些协议以及与 SpaceX 的合作，尽管其中一些协议带来的收益还需要时间才能显现。