aifollow.news Search
Back Rohan Paul
Rohan Paul· @rohanpaul_ai · X· · Original publication time SelectedAI score67

Moonshot releases Kimi K3.1 Pro after reducing its exploit capabilityMachine translation

Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.

AI introduction

Moonshot says removing cyber-related training data lowered Kimi K3.1 Pro’s pass rate on its internal exploit test from 90.5% for the unfiltered version to 81.0%, still above K3’s 33.3%. The model and a Highspeed variant running 2–3 times faster are live in Kimi apps and the API; weights are due by November 4, 2026.

Original post · AI translation

@rohanpaul_ai
Translation · refer to the original

Moonshot 在开放发布前刻意削弱了 Kimi K3.1 Pro 的黑客攻击能力。

未经削弱的版本完成了 90.5% 的内部漏洞利用任务,而发布的模型完成了 81%;按 Moonshot 的统计,这仍远高于 K3 的 33.3%,但低于 DeepSeek V4.1-Flash 的 85.7%。https://t.co/6HmruAqHXW

QuoteRohan Paul@rohanpaul_ai
Translation
Moonshot AI 的首次更新:全新 Kimi K3.1 Pro > Kimi K3.1 Pro 在 Terminal-Bench 4.0 上得分 65.7%,比 Claude Opus 5.5 低 0.7 个百分点,比 GPT-6 Astra 高 7.5 个百分点,而每项任务的平均成本低于这两个对手各自表现最佳的运行。 > Moonshot 发现,随着后训练规模扩大,漏洞利用能力也随之提高,因此从中期训练和强化学习中移除了网络安全相关数据;在其内部的 xdev-bench-hard 测试中,pass@3 从 90.5% 降至 81.0%。 > 经过筛选的模型仍能完成其中 81.0% 的漏洞利用任务,而 K3 为 33.3%。Moonshot 认为,这表明通用编程能力的提升延伸到了网络安全工作;与此同时,第三方运行的 GPT-5.6 Sol 完成了 100%。 > 在 Moonshot 的内部工具辅助幻觉压力测试中,错误断言的比例从 K3 的 23.4% 降至 8.3%;这一比例仍高于 GPT-6 Astra 的 5.5%,但远低于 Claude Opus 5.5 的 15.0%。 > 最大的跃升出现在 Moonshot 的模拟商业运营测试中,得分从 8.9 升至 37;K3.1 Pro 还以 64.4 分位居 OfficeQA Pro v2 榜首,高于 Opus 5.5 的 62.2 分。 > 提升并不均衡:K3.1 Pro 在 Terminal-Bench Science 上得分 38.1,GPT-6 Astra 和 Opus 5.5 分别为 68.1 和 63.3;但它在 Agents’ Last Exam 上以 43.2 分领先,在可使用工具的 Humanity’s Last Exam 上以 70.0 分领先。 > K3.1 Pro 及运行速度快 2–3 倍的 Highspeed 变体已在 Kimi 的应用和 API 中上线,完整权重预计于 11 月 4 日发布,供外部独立复测。
Original

Moonshot AI’s first update, the new Kimi K3.1 Pro > Kimi K3.1 Pro scores 65.7% on Terminal-Bench 4.0, 0.7 points behind Claude Opus 5.5 and 7.5 ahead of GPT-6 Astra, at a lower average cost per task than either rival’s top runs. > Moonshot saw exploit skill climb as post-training scaled, so it removed cyber-related data from mid-training and RL; on its internal xdev-bench-hard, pass@3 fell from 90.5% to 81.0%. > The filtered model still solves 81.0% of those exploitation tasks against K3’s 33.3%, which Moonshot reads as general coding gains carrying into cyber work, while a third-party run of GPT-5.6 Sol solved 100%. > On Moonshot’s internal hallucination stress test with tools, false claims fell from 23.4% for K3 to 8.3%, still above GPT-6 Astra’s 5.5% but well below Claude Opus 5.5’s 15.0%. > The biggest jump is on Moonshot’s simulated business-operation test, where scores rose from 8.9 to 37, and K3.1 Pro also tops OfficeQA Pro v2 at 64.4 against Opus 5.5’s 62.2. > Gains are uneven: K3.1 Pro scores 38.1 on Terminal-Bench Science against GPT-6 Astra’s 68.1 and Opus 5.5’s 63.3, while leading Agents’ Last Exam at 43.2 and Humanity’s Last Exam with tools at 70.0. > K3.1 Pro and a Highspeed variant running 2-3x faster are live in Kimi’s apps and API, with full weights due by November 4 for independent reruns.

View this post on X ↗
Related contextKimi.ai@Kimi_Moonshot
Translation
推出 Kimi K3.1 Pro:我们能力最强的 3T 级开放权重模型。 🔹 100 万 token 上下文窗口,参数总量 2.8T / 激活参数 104B(MoE),支持多模态。 🔹 全项目编程:使用 Unreal、Unity 和 Godot 开发游戏。 🔹 擅长在 Blender 和 CAD 中进行参数化 3D 建模。 🔹 财务工作更可靠:建模准确,幻觉减少 65%。 🔹 在编程、智能体和知识基准测试中媲美最强的专有模型。 Kimi K3.1 Pro 和 K3.1 Pro Highspeed(速度快 2–3 倍)现已在 https://t.co/85jA71X9gw、Kimi Work、Kimi Code 和 Kimi API 上线。 开放权重将于 2026 年 11 月 4 日发布。 🔗 API:https://t.co/XCrgjXAqMw 🔗 技术博客:https://t.co/5iPP0MBej6
Original

Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model. 🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal. 🔹 Full-project coding: games in Unreal, Unity & Godot. 🔹 Excels at parametric 3D modeling in Blender & CAD. 🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%. 🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks. Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/85jA71X9gw, Kimi Work, Kimi Code, and the Kimi API. Open weights by Nov 4, 2026. 🔗 API: https://t.co/XCrgjXAqMw 🔗 Tech blog: https://t.co/5iPP0MBej6

View this post on X ↗

Source: Rohan Paul · x.com

Models
Found an error? Send a correction