Introducing System One Models and Jev

介绍 System One 模型与 Jev

Diogo Almeida, founder, TypeSafe Diogo Almeida，TypeSafe 创始人

Models have been superhuman at chat for years, so where is all the automation? This has been my driving question for the last four years. At OpenAI, I helped build the methods that made language models useful at following instructions and talking with people. That work ended up as the research behind ChatGPT. At the time, I thought maybe chat models would lead to AGI, but despite the hype it became obvious to me that there was something really big missing. 多年来，模型在聊天方面已经展现出超人的能力，但自动化在哪里？这是过去四年一直驱动我思考的问题。在 OpenAI，我参与构建了使语言模型能够有效遵循指令并与人交流的方法。这些工作最终成为了 ChatGPT 背后的研究基础。当时，我曾认为聊天模型可能会通向通用人工智能（AGI），但尽管炒作不断，我逐渐意识到其中缺失了非常关键的一环。

After two years in stealth, countless technical challenges, and research breakthroughs… I am beyond excited to announce that today, TypeSafe AI is releasing our first System One Model: a new class of frontier models built to make fast, structured decisions that software can use directly. 经过两年的隐身开发、无数的技术挑战和研究突破……我非常激动地宣布，今天 TypeSafe AI 正式发布我们的首个“系统一”（System One）模型：这是一类全新的前沿模型，旨在做出软件可以直接使用的快速、结构化的决策。

We built a new stack entirely focused on automation: with a new model architecture, parallel sampler for maximum efficiency, and training method we call Reinforcement Learning for Calibrated Decisions (RLCD). 我们构建了一个完全专注于自动化的新架构栈：包括全新的模型架构、用于实现最高效率的并行采样器，以及我们称之为“校准决策强化学习”（RLCD）的训练方法。

Our first public model is Jev, available today in early access. Jev achieves similar levels of intelligence on System One tasks compared to existing LLMs, while being two orders of magnitude faster and more efficient. While Jev gives up string generation, it’s optimized for structured outputs and can’t hallucinate. Think of Jev as a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out. Extraordinary claims require extraordinary evidence so see below for the receipts. 💅 我们的首个公开模型是 Jev，即日起开放早期访问。在“系统一”任务上，Jev 达到了与现有大语言模型（LLM）相当的智能水平，同时速度和效率提升了两个数量级。虽然 Jev 放弃了字符串生成，但它针对结构化输出进行了优化，且不会产生幻觉。可以将 Jev 视为一种前沿智能函数调用：输入非结构化状态，输出类型化的概率决策。非凡的声明需要非凡的证据，请查看下方的证明。💅

---

Frontiers, Old and New

前沿：过去与现在

Feature	Existing LLMs	System One + Jev
特性	现有 LLM	System One + Jev

Optimized with	Reinforcement Learning with Human Feedback (RLHF) / Reinforcement Learning with Verifiable Rewards (RLVR)	Reinforcement Learning for Calibrated Decisions (RLCD)
优化方式	基于人类反馈的强化学习 (RLHF) / 基于可验证奖励的强化学习 (RLVR)	校准决策强化学习 (RLCD)

Optimizes for	Human preference: writeups and chat responses that human raters prefer.	Verifiable rewards: outputs that can be programmatically verified. Calibrated decisions: answers with epistemically honest probabilities on System One tasks.
优化目标	人类偏好：人类评估者更喜欢的文案和聊天回复。	可验证的奖励：可通过程序验证的输出。校准决策：在“系统一”任务中提供具有认知诚实概率的答案。

Inputs	Unstructured data (e.g. text) with an emphasis on sequential messages.	Unstructured data (e.g. text) with an emphasis on structured program state.
输入	非结构化数据（如文本），侧重于序列消息。	非结构化数据（如文本），侧重于结构化程序状态。

Outputs	Strings / generated text. Strings are flexible and can be anything: chat responses, code, hallucinations, refusals, or even type-safe structured values. To be used by software, responses need to be parsed + validated. There is also always some risk that the AI goes off the rails.	Type-safe structured values. Possible outputs and structure are defined in advance. The model never makes type errors. All answers are accompanied with calibrated probabilities and confidence scores.
输出	字符串/生成的文本。字符串非常灵活，可以是任何内容：聊天回复、代码、幻觉、拒绝，甚至是类型安全的结构化值。若要被软件使用，回复需要经过解析和验证。此外，AI 总是存在“失控”的风险。	类型安全的结构化值。可能的输出和结构均预先定义。模型绝不会出现类型错误。所有答案都附带校准后的概率和置信度分数。

Sampling	Sequential. Generates one token at a time, each conditioned on the last.	Parallel. Generates all outputs in a single query. Incredibly efficient and hardware-aware.
采样	顺序采样。一次生成一个 Token，每个 Token 都以前一个为条件。	并行采样。在单次查询中生成所有输出。极其高效且具备硬件感知能力。

Cost	Input tokens: from $0.20 to $10 / MTok. Output tokens: ~5x more expensive than input tokens.	Input tokens: $0.042 / MTok ($42 per billion tokens). Output tokens: FREE (too cheap to meter).
成本	输入 Token：$0.20 到 $10 / 百万 Token。输出 Token：比输入 Token 贵约 5 倍。	输入 Token：$0.042 / 百万 Token（每十亿 Token $42）。输出 Token：免费（成本低至无需计量）。

Speed	End-to-end response time is 3 to 329 seconds for frontier models. Fast enough for interfacing with humans, but a big bottleneck when integrated in code.	End-to-end response time is 70ms-500ms for TypeSafe. This can range from 40x-200x faster for the same levels of frontier intelligence for System One shaped queries.
速度	前沿模型的端到端响应时间为 3 到 329 秒。对于人机交互足够快，但在集成到代码中时是一个巨大的瓶颈。	TypeSafe 的端到端响应时间为 70ms-500ms。对于“系统一”类型的查询，在同等前沿智能水平下，速度可提升 40 到 200 倍。

Confidence	Even if prompted for a confidence estimate, models tend to be overconfident and inconsistent. If a model can do a task 95% of the time but doesn’t say when it’s in the 5%, it can’t automate that task.	Always communicates confidence and uncertainty with every output. Calibrated: higher confidence means higher accuracy. More consistent: returns similar answers for similar inputs.
置信度	即使被要求提供置信度估计，模型往往也过于自信且不一致。如果一个模型在 95% 的情况下能完成任务，但无法说明何时处于那 5% 的失败区间，它就无法实现该任务的自动化。	每次输出都始终传达置信度和不确定性。已校准：置信度越高意味着准确率越高。更一致：对相似的输入返回相似的答案。

Use cases	Human-in-the-loop tasks (chatbots, copilots, coding agents). Verifiable problems (math proofs, kernel optimization). Demos.	AI-Powered Workflows / smart if-statements. Map-reducing over big data. Real-time applications. Verify everything.
用例	人机协作任务（聊天机器人、副驾驶、编程代理）。可验证问题（数学证明、内核优化）。演示。	AI 驱动的工作流 / 智能 if 语句。大数据 Map-Reduce。实时应用。验证一切。