成立于去年的 AI 初创公司 Recursive 正开始名副其实。该公司致力于构建能够自我改进的 AI，于周三发布了其首个模型 Flint。Flint 是一款免费的编程模型，可在笔记本电脑上运行，且主要在 AI 的帮助下构建而成。
Recursive, an AI startup founded last year, is starting to live up to its name. The company, which is trying to build AI that can improve itself, released its first model, Flint, on Wednesday. Flint is a free coding model that can run on a laptop and was built largely with the help of AI.
Recursive 表示，其自动化研究系统——称为“尤里卡机器”（Eureka Machine）——协助了 Flint 开发的几乎每一个步骤。这使得一个小团队能在不到四个月的时间里从零开始构建该模型。
Recursive’s automated research system, which it calls the Eureka Machine, helped with nearly every step of Flint’s development, the company says. That let a small team build the model from scratch in under four months.
Flint 是“递归自我改进”这一理念的早期公开测试：即 AI 可以构建更好的 AI，而后者又能构建更好的 AI。这也是业界争议最大的想法之一。一些研究人员担心，AI 自我改进可能会脱离人类控制。目前，Recursive 的版本更为有限：在人类监督下的 AI 研究助手，产出了一个足够小、可在笔记本电脑上运行的模型。
Flint is an early public test of recursive self-improvement: the idea that AI can build better AI, which in turn builds even better AI. It’s also one of the most debated ideas in the industry. Some researchers fear that AI improving itself could slip out of human control. For now, Recursive’s version is more limited: an AI research assistant overseen by people that produced a model small enough for a laptop.
该公司于 5 月结束保密状态亮相，获得包括 GV、英伟达和 AMD Ventures 在内的投资方注资 6.5 亿美元，估值达 46.5 亿美元。Recursive 由首席执行官 Richard Socher 领导，他曾任 Salesforce 首席科学家。
The company emerged from stealth in May with $650 million in funding at a $4.65 billion valuation, from backers including GV, Nvidia, and AMD Ventures. Recursive is led by CEO Richard Socher, Salesforce’s former chief scientist.
由于 Flint 运行在用户自己的电脑上，而非公司的云端服务器，开发者可离线使用、保持代码私密，并避免 AI 云服务按模型处理文本量收取的使用费。该模型可在热门 AI 模型分享网站 Hugging Face 上免费下载，采用 Apache 2.0 许可证，这是一种允许免费使用和修改的开源许可证。
Because Flint runs on a user’s own computer rather than a company’s cloud servers, developers can use it offline, keep their code private, and avoid the usage fees that AI cloud services charge based on how much text the model processes. The model is free to download on Hugging Face, the popular site for sharing AI models, under the Apache 2.0 license, an open-source license that allows free use and modification.
Flint 是一种专家混合模型，意味着它在处理任何特定任务时仅使用自身的一小部分：其 400 亿个设置（即参数）中仅 25 亿个。这使其轻量到足以在普通笔记本电脑上运行，且远小于 Meta 上周发布的为 Mac 打造的开源模型 Muse Spark Open。后者使用 8160 亿参数，至少需要 1 TB 内存。
Flint is a mixture-of-experts model, meaning it uses only a small slice of itself for any given task: just 2.5 billion of its 40 billion settings, known as parameters, at a time. That makes it light enough for an ordinary laptop and far smaller than Meta’s Muse Spark Open, an open model built for Macs that was released last week. Muse Spark Open uses 816 billion parameters and needs at least a terabyte of memory.
Flint 速度也很快。在 Recursive 的测试中，它在搭载苹果 M4 Pro 芯片的 MacBook Pro 上每秒生成约 131 个 token（文本块）——大致是同类模型在同一台机器上 58 到 69 个 token/秒的两倍，这些同类模型包括 Google 的 Gemma 4 和 Nvidia 的 Nemotron 3.5 Lightning。在 Nvidia 的 RTX Spark 上，Flint 的速度超过每秒 150 个 token。
Flint is also fast. In Recursive’s tests, it generated about 131 tokens (chunks of text) per second on a MacBook Pro with Apple’s M4 Pro chip — roughly twice the 58 to 69 tokens per second that similar models managed on the same machine, including Google’s Gemma 4 and Nvidia’s Nemotron 3.5 Lightning. On Nvidia’s RTX Spark, Flint topped 150 tokens per second.
在 SWE-Bench Verified 基准测试中，该测试衡量 AI 修复真实软件漏洞的能力，Flint 的表现优于其他参数量低于 500 亿的模型。在 Recursive 自身跨六项测试的平均得分中，仅有一个模型——阿里巴巴的 Qwen3.8 27B——得分高于 Flint。
On SWE-Bench Verified, a benchmark that measures how well AI can fix real-world software bugs, Flint outperformed other models under 50 billion parameters. On Recursive’s own average across six tests, only one model, Alibaba’s Qwen3.8 27B, scored higher than Flint.
公司承认，与大型付费编程助手相比，Flint 需要更多指引，并建议将大型任务拆解为更小的步骤。尽管如此，早期评价是积极的。《Wired》的 Will Knight 在一台配备 48GB 内存的 Mac 上使用 Flint 构建了几个小程序，包括一个待办事项应用和一个在网络上搜寻研究资料的工具。他发现，较小的编程模型经常报错或运行缓慢，相比之下 Claude 或 Codex 表现更好，但 Flint 每次都精准完成了他的编程需求。
The company acknowledged that Flint would need more guidance than the big paid coding assistants, and suggested breaking large tasks into smaller steps. Still, early reviews were positive. Wired’s Will Knight used Flint on a Mac with 48GB of memory to build several small programs, including a to-do list app and a tool that scours the web for research. He found that smaller coding models often ran into errors or were slow compared to Claude or Codex, but Flint nailed his coding requests every time.
Eureka Machine 能够提出想法、运行实验，并利用结果决定下一步尝试什么。根据公司发布的视频，它几乎参与了构建 Flint 的每一个环节，从预训练到后训练再到推理。
The Eureka Machine comes up with ideas, runs experiments and uses the results to decide what to try next. According to the company’s launch video, it contributed to nearly every part of building Flint, from pre-training to post-training and inference.
Recursive 表示，Eureka 的贡献包括更稀疏的模型结构、低比特训练使 Flint 体积缩减 71% 且质量几乎无损，以及定制内核——这些是加速底层计算的专用程序。
Recursive said that Eureka’s contributions include a sparser model shape, low-bit training that shrank Flint by 71% with almost no loss in quality and custom kernels, which are specialized programs that speed up the underlying computations.
Recursive 未透露 Eureka Machine 为研究人员节省了多少时间，但它并非唯一依靠 AI 进行研究的公司。OpenAI 上个月表示，到 8 月中旬，其 AI 智能体每记录 3.1 个工作日，人类研究人员仅工作 1 个工作日。
Recursive didn’t say how much time the Eureka Machine saved its researchers, but it’s not alone in leaning on AI to do research. OpenAI said last month that by mid-August, its AI agents were logging 3.1 workdays for every one worked by a human researcher.
如果 Recursive 的押注奏效，每个新模型都应能在无需人工帮助的情况下承担更长时间的工作。该公司此前曾告诉《科技博客》（TechCrunch），其于 6 月与亚马逊云科技（Amazon Web Services）达成的 4.1 亿美元算力交易“与人员数量关系不大，更多是关于智能体数量”。索切尔对《连线》（Wired）表示，长远目标是让“尤里卡机器”实现完全的递归自我改进，由人类监督整个过程，并在出现问题时能够叫停。
If Recursive’s bet pays off, each new model should be able to take on longer stretches of work without human help. The company previously told TechCrunch that its $410 million compute deal with Amazon Web Services in June was “less about headcount and more about agent count.”The longer-term goal, Socher told Wired, is for the Eureka Machine to achieve full recursive self-improvement, with humans overseeing the process and able to stop it if something goes wrong.
Recursive 称，每个新模型都应让“尤里卡机器”成为更优秀的研究者，从而帮助它更快地构建下一个模型。
Recursive says each new model should make the Eureka Machine a better researcher, helping it build the next one faster.