Recursive launches Flint, a laptop coding model it says was largely built by AI

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’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.

去年成立的 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，并于周三发布了其首个模型 Flint。Flint 是一款可以在笔记本电脑上运行的免费编码模型，其构建过程很大程度上得益于 AI 的帮助。据该公司称，其名为“尤里卡机器”（Eureka Machine）的自动化研究系统为 Flint 的开发提供了几乎每个步骤的协助，这使得一个小团队能在不到四个月的时间内从零开始构建该模型。

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.

Flint 是递归自我改进这一概念的早期公开测试：即 AI 可以构建更好的 AI，进而构建出更强大的 AI。这也是业内争议最大的想法之一。一些研究人员担心，AI 的自我改进可能会脱离人类的控制。目前，Recursive 的版本还比较有限：它是一个由人类监督的 AI 研究助手，旨在生成一个足够小、能在笔记本电脑上运行的模型。

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. 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.

该公司于五月结束隐身状态，获得了 6.5 亿美元的融资，估值达到 46.5 亿美元，投资方包括 GV、英伟达和 AMD Ventures。Recursive 由前 Salesforce 首席科学家 Richard Socher 担任 CEO。由于 Flint 在用户自己的电脑上运行，而非公司的云服务器，开发者可以离线使用它，保持代码私密性，并避免 AI 云服务根据模型处理文本量收取的费用。

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 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.

该模型可在热门 AI 模型共享网站 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协议。Flint 是一款混合专家模型，意味着它在执行任何给定任务时只使用自身的一小部分：在其 400 亿个参数设置中，每次仅调用 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 上周发布的专为 Mac 构建的开源模型 Muse Spark Open。

Muse Spark Open uses 816 billion parameters and needs at least a terabyte of memory. 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.

Muse Spark Open 使用 8160 亿个参数，至少需要 1TB 的内存。Flint 的速度也很快。在 Recursive 的测试中，它在搭载苹果 M4 Pro 芯片的 MacBook Pro 上每秒可生成约 131 个 token（文本块），大约是同类模型（包括谷歌的 Gemma 4 和英伟达的 Nemotron 3.5 Lightning）在同一台机器上每秒 58 到 69 个 token 的两倍。在英伟达的 RTX Spark 上，Flint 的速度超过了每秒 150 个 token。

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. 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.

在衡量 AI 修复现实世界软件漏洞能力的基准测试 SWE-Bench Verified 中，Flint 的表现优于其他 500 亿参数以下的模型。在 Recursive 自身的六项测试平均得分中，仅有阿里巴巴的 Qwen3.8 27B 模型得分高于 Flint。该公司承认，与大型付费编码助手相比，Flint 需要更多的引导，并建议将大型任务拆解为更小的步骤。尽管如此，早期的评价依然正面。

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.

《连线》杂志的 Will Knight 在一台拥有 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用于搜索网络研究资料的工具。他发现，与 Claude 或 Codex 相比，较小的编码模型往往容易出错或运行缓慢，但 Flint 每次都能精准完成他的编码请求。

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 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.

“尤里卡机器”负责提出想法、运行实验，并利用结果决定下一步的尝试方向。根据该公司的发布视频，它参与了构建 Flint 的几乎每个环节，从预训练到后训练再到推理。Recursive 表示，Eureka 的贡献包括更稀疏的模型结构、使 Flint 体积缩小 71% 且几乎不损失质量的低位训练，以及用于加速底层计算的专用程序——自定义内核。

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 workday worked by a human researcher. If Recursive’s bet pays off, each new model should be able to take on longer stretches of work without human help.

Recursive 没有透露“尤里卡机器”为研究人员节省了多少时间，但依靠 AI 进行研究并非其独有。OpenAI 上个月表示，截至 8 月中旬，其 AI 智能体每工作 3.1 个工作日，人类研究员仅工作 1 个工作日。如果 Recursive 的赌注成功，每一个新模型都应该能够在没有人类帮助的情况下承担更长时间的工作。

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 says each new model should make the Eureka Machine a better researcher, helping it build the next one faster.

该公司此前曾告诉 TechCrunch，其 6 月份与亚马逊云科技达成的 4.1 亿美元计算协议“与其说是为了增加员工人数，不如说是为了增加智能体数量”。Socher 对《连线》表示，长远目标是让“尤里卡机器”实现完全的递归自我改进，同时由人类监督整个过程，并在出现问题时能够随时叫停。Recursive 表示，每一个新模型都应使“尤里卡机器”成为更好的研究者，从而帮助它更快地构建下一个模型。