夜雨聆风
OpenAI-backed legal tech firm pivots to Chinese Kimi K3 open-weight model
San Francisco-based Harvey says its new model, Harvey Tenet, was post-trained on top of the open-weight Kimi K3 base
OpenAI支持的法律科技公司转向中国Kimi K3开源权重模型总部位于旧金山的Harvey表示，其新模型Harvey Tenet是在开源权重模型Kimi K3基础之上进行后训练而成的
SCMP| Aug 2026
A US artificial intelligence start-up backed by OpenAI has built its first in-house model on Chinese lab Moonshot AI’s Kimi K3, highlighting a growing shift by Western tech firms towards Chinese open-weight systems amid soaring development costs.
一家人工智能初创公司获得了OpenAI的支持，并已在中国实验室月之暗面（Moonshot AI）的Kimi K3基础之上构建了其首个内部模型，这凸显出在开发成本飙升之际，西方科技公司正日益转向中国的开源权重系统。
San Francisco-based legal tech provider Harvey, whose high-profile backers also include Sequoia Capital and Andreessen Horowitz, said on Thursday that its new model, Harvey Tenet, was post-trained on top of the open-weight Kimi K3 base.
总部位于旧金山的法律科技提供商Harvey周四表示，其新模型Harvey Tenet是在开源权重模型Kimi K3基础之上进行后训练而成的。该公司的知名支持者还包括红杉资本（Sequoia Capital）和安德森·霍洛维茨（Andreessen Horowitz）。
The company said the system achieved “state-of-the-art” performance in complex legal work. The release marks a significant departure for Harvey, which serves major international law firms and enterprise clients. The start-up previously focused on customising closed proprietary models from US leaders such as Anthropic,OpenAI and Google for legal applications.
该公司表示，该系统在复杂的法律工作中达到了“最先进”的水平。这一发布标志着Harvey的一次重大转变。这家初创公司服务于大型国际律师事务所和企业客户，此前专注于为法律应用定制Anthropic、OpenAI和谷歌等美国领先企业的闭源专有模型。
Harvey’s pivot was “a great example of open-weight models” enabling developers to post-train systems on specific industry or corporate data for higher accuracy and lower inference costs, AI policy researcher Simon Hedlin wrote on social media on Friday.
AI政策研究员西蒙·赫德林（Simon Hedlin）周五在社交媒体上写道，Harvey的转向是“开源权重模型的一个绝佳范例”，它让开发者能够针对特定行业或企业数据进行后训练，从而提高准确性并降低推理成本。
Post-training is the process of refining a general-purpose base model with specialised data sets to excel at specific tasks.
后训练是指利用专门的数据集对通用基础模型进行优化，使其在特定任务上表现更出色的过程。
“We’re still only in the very earliest stages of exploring what’s possible to do with highly capable open-weight models,” Hedlin said.
“对于能力强大的开源权重模型，我们仍处于探索其可能性的最初阶段。”赫德林说。
“遗憾的是，美国在开发前沿开源权重模型方面正落后于人。”
“It’s unfortunate that America is lagging behind in developing frontier open-weight models.”
Founded in the summer of 2022, the start-up reached a valuation of US$11 billion in a financing round in March.
这家初创公司成立于2022年夏，在3月的一轮融资中估值达到110亿美元。
Harvey said in a blog post on Thursday that its research over the past six months focused on building “frontier legal intelligence using open-weight models” and enabling law firms to build and deploy specialised models.
Harvey在周四的一篇博客文章中表示，过去六个月其研究重点是利用开源权重模型构建“前沿法律智能”，并帮助律师事务所构建和部署专用模型。
Trained on comprehensive legal data sets, Harvey Tenet outperformed both its underlying base model and US frontier systems – including Fable 5 and GPT-5.6 Sol – across a range of complex, long-horizon legal agentic tasks, according to the company.
据该公司介绍，Harvey Tenet在全面的法律数据集上训练而成，在一系列复杂的、长周期的法律智能体任务中，表现优于其底层基础模型以及美国的前沿系统——包括Fable 5和GPT-5.6 Sol。
The approach also improved cost efficiency, according to Harvey. The firm said that while open-weight models naturally offered lower per token prices, it also worked on reducing the number of tokens consumed in inference.
Harvey表示，这种方法还提高了成本效率。该公司称，虽然开源权重模型天然提供更低的每词元价格，但它还致力于减少推理过程中消耗的词元数量。
The training for Harvey Tenet was done over two months using around 150 Nvidia B300 graphics processing units (GPUs), the company said.
该公司表示，Harvey Tenet的训练历时两个月，使用了约150块英伟达B300图形处理器（GPU）。
Open-source models such as Nvidia’s Nemotron series now account for 40 per cent of US telecommunications giant AT&T’s employee AI queries, according to a Thursday report by The Information, citing AT&T vice-president Mark Austin.
据科技媒体The Information周四报道，目前开源模型（如英伟达的Nemotron系列）已占美国电信巨头AT&T员工AI查询量的40%。该报道援引了AT&T副总裁马克·奥斯汀（Mark Austin）的说法。
Austin reportedly also said that AT&T was not currently using any Chinese open-weight models.
据报道，奥斯汀还表示AT&T目前并未使用任何中国的开源权重模型。
The company was still “evaluating the potential risks” of using them, he reportedly said, and analysing options from Chinese firms including DeepSeek and Moonshot, the Beijing-based company behind Kimi K3.
....
重点词汇解析：
1. pivot /ˈpɪvət/ v. 战略转向，转型
2. back /bæk/ v. 投资，资助（初创公司）
3. legal‑tech n. 法律科技
4. high‑profile adj. 知名度高的，大牌的
5. start‑up n. 初创企业
6. valuation /ˌvæljuˈeɪʃn/ n. 估值
7. enterprise client 企业客户
9. open‑weight model 开放权重模型
10. base model 底座基础模型
...
文末测验：
1. What major move has Harvey made according to the article?
A. It shut down all AI‑related business completely
B. It built its first in‑house model Harvey Tenet post‑trained on China’s Kimi K3 open‑weight base model
C. It only keeps using closed US models and rejects open‑weight systems
D. It was acquired by Moonshot AI of China
2. What is post‑training in AI mentioned in this text?
A. Only training brand‑new model from zero
B. Refining a general‑purpose base model with specialized datasets for specific tasks
C. Purely human manual editing of model output texts
D. Hardware manufacturing for GPU chips
3. According to Harvey, one benefit brought by open‑weight‑based approach is ______.
A. Zero hardware consumption
B. Improved cost efficiency including lower inference‑related token consumption
C. Completely free from all potential risks
以上为原文节选，欲阅读全文，欢迎购买我们的外刊精读专栏（夏季焕新价！！最新鲜的第一手资讯，你值得拥有！)；或者通过我们的同名zhihu获取单篇阅读~
下载app后即扫一扫即可加入阅览精读营所有文章~购买了的朋友别忘了私信加粉丝群哦~
——定期精选优质外刊精读文章、经典书籍，文章精选，话题丰富且与时俱进，契合考研、托福\雅思、GRE\GMAT等考试选题，以期帮助大家打破思维桎梏，发现未知的自己~
每日精选优质外刊文章
做你英语学习路上的好搭子!