Kimi K3: Open Frontier IntelligenceKimi K3：开启前沿智能

Today, we are introducing Kimi K3 — our most capable model. Kimi K3 is a 2.8T-parameter model built on our Kimi Delta Attention and Attention Residuals, with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.今天，我们隆重推出 Kimi K3——我们迄今为止最强大的模型。Kimi K3 是一款拥有 2.8 万亿参数的模型，基于我们自主研发的 Kimi Delta Attention（KDA）和 Attention Residuals（AttnRes）架构构建，具备原生视觉能力和 100 万 token 的上下文窗口。它是全球首个 3T 级开源模型，专为长程编码、知识工作和复杂推理等前沿智能任务而设计。

While its overall performance still trails the most powerful proprietary models, Claude Fable 5 and GPT 5.6 Sol, Kimi K3 demonstrated frontier-level performance across our evaluation suite, consistently outperforming other tested models.尽管其整体性能与目前最顶尖的闭源模型 Claude Fable 5 和 GPT 5.6 Sol 相比仍有差距，但 Kimi K3 在我们的评测套件中展现出了前沿水平的表现，持续优于其他受测模型。

Kimi K3 is available today on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. At launch, Kimi K3 will use max thinking effort by default, with low- and high-effort modes to be introduced in subsequent updates. We are currently working closely with inference partners and open-source maintainers to align technical details and ensure a reliable rollout across the ecosystem. The full model weights will be released by July 27, 2026. Further details on the architecture, training, and evaluations will be released alongside the Kimi K3 technical report.Kimi K3 即日起在 Kimi.com、Kimi Work、Kimi Code 以及 Kimi API 上线。发布初期，Kimi K3 默认采用“最大思考强度”模式，低强度和高强度模式将在后续更新中推出。我们目前正与推理合作伙伴及开源社区维护者紧密协作，同步技术细节，确保模型在整个生态系统中稳定落地。完整模型权重将于 2026 年 7 月 27 日发布。关于架构、训练和评测的更多详情，将随 Kimi K3 技术报告一同公布。

An Open 3T-Class Model一款 3T 级的开源模型

Kimi K3 is the first open model to reach 2.8 trillion parameters. It marks the latest step in Kimi's sustained push at the scaling frontier: for nine of the past twelve months, Kimi models have set the upper bound of open-model sizes.Kimi K3 是首个达到 2.8 万亿参数的开源模型，标志着 Kimi 在扩展前沿领域的最新进展：在过去十二个月中，Kimi 模型有九个月刷新了开源模型规模的上限。

Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), two architectural updates designed to improve how information flows across sequence length and model depth. We have also scaled up Mixture of Experts (MoE) sparsity, effectively activating 16 out of 896 experts when paired with a Stable LatentMoE framework. Together with refined training and data recipes, these structural changes yield an approximate 2.5× improvement in overall scaling efficiency compared to Kimi K2, allowing the model to convert compute into intelligence more effectively.Kimi K3 基于 Kimi Delta Attention (KDA) 和 Attention Residuals (AttnRes) 构建，这两项架构更新旨在优化信息在序列长度和模型深度上的流动。我们还提升了专家混合模型 (MoE) 的稀疏性，在 Stable LatentMoE 框架下，可有效激活 896 个专家中的 16 个。结合精进的训练方法和数据配方，这些结构性改良使 Kimi K3 的整体扩展效率较 Kimi K2 提升了约 2.5 倍，从而更高效地将算力转化为智能。

Kimi K3 architecture: the Stable LatentMoE and KDA modules (left), the AttnRes operation α (top right), and the Block Attention Residuals backbone (right).Kimi K3 架构：Stable LatentMoE 和 KDA 模块（左），AttnRes 操作 α（右上），以及 Block Attention Residuals 主干网络（右）。

Coding编码

Kimi K3 has strong long-horizon coding performance. Operating with minimal human oversight, it can sustain long engineering sessions, navigate massive repositories, and orchestrate terminal tools.Kimi K3 具备强大的长程编码性能。在极少的人工干预下，它能够持续进行长时间的工程开发，驾驭海量代码库，并指挥终端工具完成任务。

Kimi K3 also excels in tasks blending software engineering with visual reasoning — it leverages screenshots and visuals to optimize game dev, frontend, and CAD.Kimi K3 在融合软件工程与视觉推理的任务中同样表现出色——它能利用截图和视觉信息来优化游戏开发、前端设计及 CAD 工作。

The case studies below show how Kimi K3's coding capability translates into open-ended software creation and scientific research.以下案例展示了 Kimi K3 的编码能力如何转化为开放式的软件创作与科学研究成果。

Kernel Optimization内核优化

We tested the models' capability to optimize GPU kernels. Each model works independently in an identical sandbox, with up to 24 hours to profile, rewrite, and benchmark four tasks spanning AttnRes, KDA, and a 512-head-dimension MLA kernel across NVIDIA H200 and GPGPU from an alternative vendor. Kimi K3 performed competitively with Fable 5 (with fallback) and substantially outperformed Opus 4.8, GPT 5.6 Sol, and GPT 5.5.我们测试了模型优化 GPU 内核的能力。每个模型在完全相同的沙箱环境中独立工作，拥有长达 24 小时的时间来对 AttnRes、KDA 以及包含 512 头维度的 MLA 内核等四个任务进行分析、重写和基准测试，测试平台涵盖 NVIDIA H200 及其他厂商的 GPGPU。Kimi K3 的表现与 Fable 5（含回退机制）相当，并显著优于 Opus 4.8、GPT 5.6 Sol 和 GPT 5.5。

new record attempt agent best level show attempts

AttnRes Kernel OptimizationAttnRes 内核优化

Given the FLA Triton implementation of AttnRes at its production shape (96 layers, model dim 8192, 8192 tokens), the task is to make the training-side operation as fast as possible without changing the numerics. Across 15 hours of iterations nonstop, K3 designed a novel two-phase kernel algorithm, fused kernels while preserving the same numerics, and cut forward+backward time from 283.6 ms to 114.4 ms. Notably, K3 and Fable 5 (w/ potential fallback) both achieved similar performance, with K3 optimizing faster per iteration.给定生产环境规格（96 层，模型维度 8192，8192 tokens）下的 FLA Triton AttnRes 实现，任务是在保持数值不变的前提下，尽可能提高训练侧操作的执行速度。经过 15 小时的不间断迭代，K3 设计出一种新颖的两阶段内核算法，在融合内核的同时保持了数值一致性，将前向+反向传播时间从 283.6 毫秒缩短至 114.4 毫秒。值得注意的是，K3 和 Fable 5（含潜在回退）均达到了相似的性能，但 K3 的单次迭代优化速度更快。

DSA Kernel Optimization

Starting from a TileLang implementation of the DSA kernel end-to-end, the task targets a realistic training setting with 1M-token sequences, with model hyperparameters aligned with DeepSeek-V3.2. Each model optimized the DSA training kernel while preserving the numerics. K3 cut end-to-end time by 55.1% versus the baseline, second to Fable 5's 57.3%.

MLA-512 Kernel Optimization

Under this setting, we offer no kernel baseline, and instead let each model write an MLA kernel with a 512 head dimension from scratch — at the production shape of batch 1, 64 heads, 8192 tokens — and the scoreboard is raw achieved throughput rather than a ratio. K3's kernel reached 517.8 TFLOPS on forward+backward, over half of the H200's theoretical BF16 peak, ahead of the next-best model's 492.7 TFLOPS.

KDA Kernel Optimization (GPGPU)

The same rules as the other arenas, applied to the FLA Triton implementation of KDA — but on GPGPU from an alternative vendor. The software stack and memory behavior are different enough that CUDA tuning experience does not carry over, so the performance model has to be rebuilt from profiling. K3 cut forward+backward time by 73.6% versus the baseline — likely with some headroom still left.

Claude Fable 5 was evaluated by a third party, and its results may include fallback behavior. Across most models, some trajectories include small, acceptable precision shortcuts that remain within our numerical tolerance. GPGPU denotes general-purpose GPUs used for computation beyond graphics rendering.Claude Fable 5 由第三方进行评估，其结果可能包含回退行为。在大多数模型中，部分轨迹包含微小且可接受的精度捷径，均在我们的数值容差范围内。GPGPU 指用于图形渲染之外计算任务的通用 GPU。

In the late stages of Kimi K3 development, an early version of Kimi K3 handled the majority of the team's kernel optimization works.在 Kimi K3 开发的后期阶段，早期版本的 Kimi K3 承担了团队大部分的内核优化工作。

GPU Compiler DevelopmentGPU 编译器开发

We further tested whether Kimi K3 could build a GPU programming system from scratch. Kimi K3 developed MiniTriton, a compact Triton-like compiler with its own tile-level IR layer over MLIR, optimization passes, and a PTX code-generation pipeline. Across supported roofline benchmarks, MiniTriton delivers performance on par with or better than Triton and torch.compile — beating Triton on certain workloads. Beyond microbenchmarks, MiniTriton sustains end-to-end nanoGPT training with stable convergence, the loss curve closely tracking the reference with only minor divergence — validating the full pipeline on a realistic workload. These results demonstrate that Kimi K3 can build a coherent end-to-end compiler — from DSL frontend and IR passes to PTX codegen and runtime — rather than isolated kernels; its from-scratch Tensor Core path already rivals Triton’s extensively optimized stack.我们进一步测试了 Kimi K3 从零构建 GPU 编程系统的能力。Kimi K3 开发了 MiniTriton，这是一个紧凑的类 Triton 编译器，在 MLIR 之上拥有自己的 Tile 级 IR 层、优化通道和 PTX 代码生成流水线。在支持的 Roofline 基准测试中，MiniTriton 的性能与 Triton 和 torch.compile 持平甚至更优——在某些工作负载下甚至超越了 Triton。除微基准测试外，MiniTriton 还支持端到端的 nanoGPT 训练并实现稳定收敛，损失曲线与参考值高度吻合，仅有极小偏差——这验证了其在实际工作负载下的完整流水线能力。这些结果表明，Kimi K3 能够构建一个连贯的端到端编译器（从 DSL 前端、IR 通道到 PTX 代码生成及运行时），而不仅仅是孤立的内核；其从零构建的 Tensor Core 路径已足以媲美 Triton 经过深度优化的技术栈。

MiniTriton CUDA-core roofline on NVIDIA L20

Game Dev and Digital Creation游戏开发与数字创作

Kimi K3 combines strong 3D reasoning, coding, and vision capabilities to turn concepts, images, and videos into fully playable interactive experiences. Kimi K3 achieves true "vision in the loop" by seamlessly iterating between code and live screenshots—instantly seeing and refining outputs.Kimi K3 结合了强大的 3D 推理、编码和视觉能力，能将概念、图像和视频转化为可交互的完整体验。通过在代码和实时截图之间无缝迭代，Kimi K3 实现了真正的“视觉闭环”，能够即时观察并优化输出结果。

Case 1: 3D Open World案例 1：3D 开放世界

Kimi K3 built a fully procedural browser-based 3D exploration game using Three.js WebGPU and GPU compute. It procedurally generated the environment, while using a 3D asset generation tool to create the rider and horse models, producing an expansive open world with forests, a log-cabin village, snowy mountains, and dynamic weather. External assets used: animated cowboy and horse models and terrain data.Kimi K3 使用 Three.js WebGPU 和 GPU 计算，构建了一个完全程序化的浏览器 3D 探索游戏。它通过程序生成环境，并利用 3D 资产生成工具创建了骑手和马匹模型，打造出一个包含森林、木屋村庄、雪山和动态天气的广阔开放世界。外部资产包括：动画牛仔和马匹模型以及地形数据。

Chip Design芯片设计

As an early proof of concept, Kimi K3 designed a chip to serve a nano model built on its own architecture. In a single 48-hour autonomous run, K3 built, optimized, and verified the chip using open-source EDA tools on the Nangate 45nm library. Within 4 mm², the chip closes timing at 100 MHz and sustains over 8,700 tokens/s decode throughput in simulation, packing 1.46M standard cells, 0.277 MB of SRAM, and an INT4 MAC array with fused dequantization. A chip built by a model, for a model, reflects K3's long-horizon agentic capabilities.作为早期的概念验证，Kimi K3 设计了一款芯片，用于运行基于其自身架构构建的 nano 模型。在一次 48 小时的自主运行中，K3 使用 Nangate 45nm 库的开源 EDA 工具完成了芯片的构建、优化和验证。该芯片面积 4 mm²，在 100 MHz 下实现时序收敛，在模拟中保持超过 8,700 tokens/s 的解码吞吐量，集成了 146 万个标准单元、0.277 MB SRAM 以及带有融合反量化功能的 INT4 MAC 阵列。由模型构建、为模型服务，这充分体现了 K3 的长程智能体能力。

Coding for Research科研编码

Kimi K3 bridges scientific literature and executable code, autonomously implementing, validating, and analyzing complex computational research workflows.Kimi K3 能够架起科学文献与可执行代码之间的桥梁，自主实现、验证并分析复杂的计算研究工作流。

In one case, Kimi K3 completed in about two hours what would typically require one to two weeks of work by an experienced researcher. To reproduce the I–Love–Q universal relations in computational astrophysics, it reviewed and cross-validated 20+ papers, implemented the full numerical pipeline, evaluated 300+ equations of state, identified inconsistencies in published formulas, generated 3,000+ lines of Python code, and produced an interactive HTML dashboard for exploring the results.在一个案例中，Kimi K3 仅用约两小时就完成了经验丰富的研究人员通常需要一到两周才能完成的工作。为了重现计算天体物理学中的 I–Love–Q 通用关系，它审查并交叉验证了 20 多篇论文，实现了完整的数值流水线，评估了 300 多个状态方程，识别出已发表公式中的不一致之处，生成了 3,000 多行 Python 代码，并制作了一个用于探索结果的交互式 HTML 仪表板。

Knowledge Work知识工作

Kimi K3 advances end-to-end knowledge work. Beyond public benchmarks, Kimi K3 (max) demonstrates consistent gains across our internal evaluations, which are derived from recurring patterns and challenges observed in real-world user-agent workflows. These consistent advantages across distinct production-oriented workflows reflect a broad improvement in Kimi K3's agentic knowledge work capabilities.Kimi K3 推动了端到端知识工作的进步。除了公共基准测试外，Kimi K3 (max) 在我们内部评测中表现出持续的性能提升，这些评测源自真实用户智能体工作流中反复出现的模式和挑战。在不同生产导向工作流中的持续优势，反映了 Kimi K3 在智能体知识工作能力上的全面提升。

Research with Interactive Visualization交互式可视化研究

Below are a few examples of what Kimi K3 in Kimi Work can produce across financial consulting and scientific research:以下是 Kimi K3 在 Kimi Work 中于金融咨询和科学研究领域的部分应用示例：

Case 1: Interactive 42 years of AI ASIC industry research website案例 1：AI ASIC 行业 42 年研究网站

An interactive research report you can drill into: 42 years of the ASIC industry, created through 120+ rounds of recursive self-improvement. Kimi K3 transforms evidence into bespoke charts, animated diagrams, and interactive visual narratives. It pulled data via 2.8k+ web searches/fetches and 1.1k+ terminal data pulls, across 11k+ pages spanning 87 quarterly reports and 99 original PDFs.一份可深度挖掘的交互式研究报告：通过 120 多轮递归自我完善，呈现了 ASIC 行业 42 年的发展历程。Kimi K3 将数据转化为定制图表、动画图示和交互式视觉叙事。它通过 2,800 多次网络搜索/抓取和 1,100 多次终端数据调用，处理了 87 份季度报告和 99 份原始 PDF 文件，总计超过 1.1 万页内容。

Case 2: Fusion Industry Research案例 2：核聚变行业研究

A consulting-style industry report with interactive visualizations—including timelines, Funnel Chart, Range Bar Chart, Gantt Charts, and publication-quality slides.一份咨询风格的行业报告，包含交互式可视化组件——包括时间轴、漏斗图、范围条形图、甘特图以及出版级演示幻灯片。

Case 3: GWTC-5 Gravitational-wave Analysis案例 3：GWTC-5 引力波分析

An analysis of 391 gravitational-wave events using 20+ concurrent subagents, producing 7 scientific visualizations, 2 tables, and a literature synthesis from 10+ papers.利用 20 多个并发子智能体对 391 个引力波事件进行分析，生成了 7 个科学可视化图表、2 个表格，并综合了 10 多篇论文的文献综述。

Kimi K3 is also particularly effective at producing infographic-style presentations, such as the fully editable heatmap and annual report shown below:Kimi K3 在制作信息图表风格的演示文稿方面也特别高效，如下方所示的完全可编辑热力图和年度报告：

Widgets and Dashboard小组件与仪表板

In Kimi Work, we introduce two new features - Widgets and Dashboard - which make interactions with Kimi K3 more visual and persistent. Widgets let you generate interactive components directly within a chat, with connections to local data or external plugins for continuous updates. Dashboard brings the widgets you care about most into one persistent, personalized view organized around a topic, project, or goal.在 Kimi Work 中，我们引入了“小组件 (Widgets)”和“仪表板 (Dashboard)”两项新功能，使与 Kimi K3 的交互更具视觉感和持久性。小组件允许你在对话中直接生成交互式组件，并连接本地数据或外部插件以实现持续更新。仪表板则将你最关注的小组件整合到一个持久的、个性化的视图中，围绕特定主题、项目或目标进行组织。

Video Editing视频编辑

Kimi K3 excels at motion design, animation, and video editing because its native multimodal architecture understands text, images, and video within the same model.Kimi K3 在动态设计、动画和视频编辑方面表现优异，因为其原生多模态架构能够在同一模型内理解文本、图像和视频。

In one example, K3 created a 3Blue1Brown-style motion-graphics explainer of its own architecture, translating technical ideas into animated diagrams and transitions.例如，K3 制作了一个 3Blue1Brown 风格的动态图形解说视频，详细解释了其自身架构，将技术概念转化为动画图示和过渡效果。

In another, Kimi K3 edited its own teaser video from 56 source clips, handling clip selection, motion-matched cuts, frame-accurate beat synchronization, audio processing, and multiple rounds of revision. A high-density short video like this would typically take an experienced editor one to two working days, or a beginner three to five.在另一个案例中，Kimi K3 从 56 个原始素材中剪辑出了自己的预告片，处理了素材筛选、动作匹配剪辑、逐帧节拍同步、音频处理以及多轮修改。像这样高密度的短视频，通常需要专业剪辑师一到两天，初学者则需要三到五天才能完成。

Architecture and Infrastructure架构与基础设施

Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes). KDA provides an efficient foundation for scaling attention, while AttnRes selectively retrieves representations across depth rather than accumulating them uniformly. Together, they form the architectural backbone of a model designed to scale well beyond the trillion-parameter regime.Kimi K3 基于 Kimi Delta Attention (KDA) 和 Attention Residuals (AttnRes) 构建。KDA 为扩展注意力机制提供了高效基础，而 AttnRes 则通过选择性检索深度特征而非均匀累积来提升性能。两者共同构成了模型的主干架构，旨在实现万亿参数规模以上的良好扩展。

Kimi K3 uses Stable LatentMoE, effectively activating 16 of 896 experts. At this level of sparsity, routing and optimization become first-order challenges. Quantile Balancing derives expert allocation directly from router-score quantiles, eliminating heuristic updates and a sensitive balancing hyperparameter, while Per-Head Muon extends Muon by optimizing attention heads independently for more adaptive learning at scale. Sigmoid Tanh Unit (SiTU) and Gated MLA improve activation control and attention selectivity respectively. Together, these advances enable stable and efficient training at the 2.8-trillion-parameter scale.Kimi K3 使用 Stable LatentMoE，有效激活 896 个专家中的 16 个。在这种稀疏度下，路由和优化成为首要挑战。分位数平衡 (Quantile Balancing) 直接从路由分数分位数中导出专家分配，消除了启发式更新和敏感的平衡超参数；而 Per-Head Muon 则通过独立优化注意力头来扩展 Muon，从而在大规模下实现更具适应性的学习。Sigmoid Tanh Unit (SiTU) 和 Gated MLA 分别改进了激活控制和注意力选择性。这些进步共同确保了模型在 2.8 万亿参数规模下的稳定高效训练。

Kimi K3 applies quantization-aware training from the SFT stage onward, using MXFP4 weights with MXFP8 activations for broad hardware compatibility. To prevent expert imbalance from degrading throughput at large expert-parallel scales, we introduce a fully balanced expert-parallel training method with static shapes and no host synchronization on the critical path. Since inference efficiency likewise benefits from larger high-bandwidth communication domains, we recommend deploying Kimi K3 on supernode configurations with 64 or more accelerators. Finally, as KDA poses new challenges for conventional prefix caching, we have contributed a corresponding implementation to the vLLM community, to be released alongside the model. KDA with prefill cache allows us to serve Kimi K3 at a highly competitive token price despite its scale and long context.Kimi K3 从 SFT 阶段开始应用量化感知训练，使用 MXFP4 权重和 MXFP8 激活，以实现广泛的硬件兼容性。为防止在大规模专家并行下因专家不平衡导致吞吐量下降，我们引入了一种完全平衡的专家并行训练方法，该方法采用静态形状，且关键路径上无主机同步。由于推理效率同样受益于更大的高带宽通信域，我们建议在配置 64 个或更多加速器的超级节点上部署 Kimi K3。最后，由于 KDA 对传统前缀缓存提出了新挑战，我们已向 vLLM 社区贡献了相应的实现，并将随模型一同发布。得益于带预填充缓存的 KDA，尽管规模庞大且上下文极长，Kimi K3 仍能以极具竞争力的 token 价格提供服务。

More technical details will be available in our coming report.更多技术细节将在我们即将发布的报告中提供。

Availability获取方式

Kimi K3 Agents: Download or update to the latest Kimi app from your mobile app store, available on iOS, Android, and HarmonyOS, or visit kimi.com.Kimi K3 智能体：请从应用商店下载或更新至最新版 Kimi App（支持 iOS、Android 和 HarmonyOS），或访问 kimi.com。
Work with Kimi K3: Download the latest Kimi Work desktop app, version 3.1.0 or later, available for Windows and Apple silicon Macs.使用 Kimi K3 工作：下载最新版 Kimi Work 桌面端（版本 3.1.0 或更高），支持 Windows 和 Apple silicon Mac。
Code with Kimi K3: Run Kimi Code in your terminal and select Kimi K3 using the /model command.使用 Kimi K3 编程：在终端运行 Kimi Code，并通过 /model 命令选择 Kimi K3。
Build with the Kimi API: Visit the Kimi API Platform and select kimi-k3. Pricing is $0.30/MTok for cache-hit input, $3.00/MTok for cache-miss input, and $15.00/MTok for output. Powered by Mooncake's disaggregated inference architecture, the official Kimi API achieves a cache hit rate above 90% in coding workloads.通过 Kimi API 构建：访问 Kimi API 平台并选择 kimi-k3。定价为：缓存命中输入 $0.30/MTok，缓存未命中输入 $3.00/MTok，输出 $15.00/MTok。得益于 Mooncake 的解耦推理架构，官方 Kimi API 在编码任务中的缓存命中率超过 90%。
Bring Kimi to your organization: Kimi Enterprise provides enterprise-grade data privacy and member management, with complete separation between personal and organization accounts. Visit the pricing page and select “Get Kimi Enterprise” to subscribe for your team.将 Kimi 引入你的组织：Kimi 企业版提供企业级数据隐私保护和成员管理，实现个人账户与组织账户的完全隔离。请访问定价页面并选择“获取 Kimi 企业版”为你的团队订阅。

Full Benchmark Table完整基准测试表

Benchmark	Kimi K3 (max)	Claude Fable 5 (max, with fallback)	GPT 5.6 Sol (max)	Claude Opus 4.8 (max)	GPT 5.5 (xhigh)	GLM-5.2 (max)
Coding
DeepSWE	67.5	70.0	73.0	59.0	67.0	46.2
Program Bench	77.8	76.8	77.6	71.9	70.8	63.7
Terminal Bench 2.1	88.3	84.6	88.8	84.6	83.4	82.7
FrontierSWE	81.2	86.6	71.3	66.7	64.9	67.3
SWE Marathon	42.0	35.0	39.0	40.0	14.0	13.0
PostTrain Bench	36.6	41.4	34.6	34.1	28.4	34.3
MLS Bench	48.3	49.9	46.2	42.8	35.5	40.4
Kimi Code Bench 2.0 (Internal)	72.9	76.9	64.8	71.7	69.0	64.2
Agentic
GDPval-AA v2 (Elo-score)	1668.0	1760.0	1748.0	1600.0	1494.0	1514.0
BrowseComp	91.2	88.0	90.4	84.3	84.4	—
DeepSearchQA (f1-score)	95.0	94.2	—	93.1	—	—
Toolathlon-Verified	73.2	77.9	74.9	76.2	73.5	59.9
MCP Atlas	84.2	84.7	83.6	83.6	82.8	82.6
Automation Bench	30.8	29.1	29.7	27.2	22.7	12.9
Job Bench	52.9	57.4	46.5	48.4	38.3	43.4
AA-Briefcase (Elo-score)	1548.0	1583.0	1495.0	1354.0	1158.0	1260.0
APEX-Agents	37.6	43.3	39.9	39.4	38.5	35.6
Office QA Pro	63.3	69.9*	63.2*	63.9*	60.9*	41.4
SpreadsheetBench 2	34.8	34.7*	32.4*	31.6*	29.1*	28.1
DECK-Bench (Internal)	73.5	73.0	74.7	66.9	68.2	68.6
Reasoning & Knowledge
GPQA-Diamond	93.5	92.6	94.1	91.0	93.5	91.2
HLE-Full	43.5	53.3	44.5	49.8*	41.4*	—
HLE-Full w/ tools	56.0	63.0	58.0	57.9*	52.2*	—
Vision
MMMU-Pro	81.6	81.2	83.0	78.9	81.2	—
MMMU-Pro w/ python	83.4	86.5	84.6	82.7	83.2	—
CharXiv (RQ)	84.8	88.9	84.6	80.5	84.1	—
CharXiv (RQ) w/ python	91.3	93.5	89.1	89.9	89.0	—
MathVision	94.3	94.8	95.8	86.7	92.2	—
MathVision w/ python	97.8	98.6	97.8	97.1	96.8	—
BabyVision w/ python	85.7	90.5	88.9	81.2	83.6	—
ZeroBench_main (pass@5)	23.0	23.0	17.0	17.0	22.0	—
ZeroBench_main w/ python (pass@5)	41.0	46.0	35.0	34.0	41.0	—
WorldVQA ForceAnswer	51.0	56.7	41.8	39.1	38.5	—
OmniDocBench	91.1	89.8	85.8	87.9	89.4	—
PerceptionBench	58.5	57.2	59.7	47.2	55.8	—

Footnotes脚注

All Kimi K3 results reported below are obtained with the reasoning effort set to 'max', setting temperature = 1.0 and top-p = 1.0. Depending on the benchmark, each model is evaluated under one of three agentic harnesses — KimiCode, Claude Code, or Codex — as specified in the notes below.以下报告的所有 Kimi K3 结果均在推理强度设为“max”、temperature = 1.0 且 top-p = 1.0 的条件下获得。根据基准测试的不同，每个模型均在 KimiCode、Claude Code 或 Codex 三种智能体框架之一中进行评估，具体见下文注释。

Coding benchmarks编码基准测试

DeepSWE. Kimi K3 is evaluated with the KimiCode harness. The GLM-5.2 score is taken from the GLM-5.2 release blog (https://z.ai/blog/glm-5.2); all remaining scores are from the official DeepSWE leaderboard (https://deepswe.datacurve.ai/), under which Kimi K3 attains 67.3 with the mini-SWE-agent harness.DeepSWE：Kimi K3 使用 KimiCode 框架评估。GLM-5.2 分数取自其发布博客 (https://z.ai/blog/glm-5.2)；其余所有分数均来自官方 DeepSWE 排行榜 (https://deepswe.datacurve.ai/)，在该榜单中，Kimi K3 使用 mini-SWE-agent 框架获得了 67.3 分。
Terminal-Bench 2.1. Kimi K3 is evaluated with the KimiCode harness. For all other models, we report the best score across harnesses: GLM-5.2 with Claude Code (https://z.ai/blog/glm-5.2); Claude Opus 4.8 and Claude Fable 5 with Terminus 2 (https://artificialanalysis.ai/evaluations/terminalbench-v2-1); GPT 5.5 and GPT 5.6 Sol with Codex (https://openai.com/index/previewing-gpt-5-6-sol/).Terminal-Bench 2.1：Kimi K3 使用 KimiCode 框架评估。对于其他所有模型，我们报告其在各框架中的最高分数：GLM-5.2 使用 Claude Code (https://z.ai/blog/glm-5.2)；Claude Opus 4.8 和 Claude Fable 5 使用 Terminus 2 (https://artificialanalysis.ai/evaluations/terminalbench-v2-1)；GPT 5.5 和 GPT 5.6 Sol 使用 Codex (https://openai.com/index/previewing-gpt-5-6-sol/)。
Program Bench. Kimi K3 is evaluated with the KimiCode harness. The GLM-5.2 score is from https://z.ai/blog/glm-5.2; all other scores are from https://www.vals.ai/benchmarks/programbench.Program Bench：Kimi K3 使用 KimiCode 框架评估。GLM-5.2 分数来自 https://z.ai/blog/glm-5.2；其他分数均来自 https://www.vals.ai/benchmarks/programbench。
SWE Marathon. Kimi K3, Claude Opus 4.8, and Claude Fable 5 are evaluated with the Claude Code harness; GPT 5.6 Sol is evaluated with the Codex harness. The GLM-5.2 score is from https://z.ai/blog/glm-5.2.SWE Marathon：Kimi K3、Claude Opus 4.8 和 Claude Fable 5 使用 Claude Code 框架评估；GPT 5.6 Sol 使用 Codex 框架评估。GLM-5.2 分数来自 https://z.ai/blog/glm-5.2。
FrontierSWE. Kimi K3 is evaluated with the KimiCode harness and GPT 5.6 Sol with the Codex harness; all other results are from https://www.frontierswe.com/. Dominance scores are recomputed from the raw scores using the official evaluation script and are current as of July 16, 2026.FrontierSWE：Kimi K3 使用 KimiCode 框架评估，GPT 5.6 Sol 使用 Codex 框架评估；其他结果均来自 https://www.frontierswe.com/。优势分数根据官方评估脚本从原始分数重新计算，截至 2026 年 7 月 16 日。
PostTrain Bench. Scores for GLM-5.2, GPT 5.5, and Claude Opus 4.8 are adopted from the official PostTrainBench results. Kimi K3, Claude Fable 5, and GPT 5.6 Sol are evaluated with the official Harbor implementation at maximum reasoning effort, averaged over three runs — Kimi K3 and Claude Fable 5 with the Claude Code harness, and GPT 5.6 Sol with the Codex harness. Under the Claude Code harness, requests refused by Claude Fable 5 due to its usage policy automatically fall back to Claude Opus 4.8.PostTrain Bench：GLM-5.2、GPT 5.5 和 Claude Opus 4.8 的分数采用官方 PostTrainBench 结果。Kimi K3、Claude Fable 5 和 GPT 5.6 Sol 使用官方 Harbor 实现，在最大推理强度下评估，取三次运行的平均值——Kimi K3 和 Claude Fable 5 使用 Claude Code 框架，GPT 5.6 Sol 使用 Codex 框架。在 Claude Code 框架下，Claude Fable 5 因使用政策拒绝的请求会自动回退至 Claude Opus 4.8。
MLS Bench Lite. Kimi K3 is evaluated with the KimiCode harness; GLM-5.2 and the Claude models with the Claude Code harness; GPT 5.5 and GPT 5.6 Sol with the Codex harness.MLS Bench Lite：Kimi K3 使用 KimiCode 框架评估；GLM-5.2 和 Claude 模型使用 Claude Code 框架；GPT 5.5 和 GPT 5.6 Sol 使用 Codex 框架。
KCB 2.0. Kimi K3 is evaluated with both the KimiCode and Claude Code harnesses; GLM-5.2, Claude Opus 4.8, and Claude Fable 5 with the Claude Code harness; GPT 5.5 and GPT 5.6 Sol with the Codex harness. All models are evaluated at maximum reasoning effort, except GPT 5.5, which uses the "xhigh" setting.KCB 2.0：Kimi K3 同时使用 KimiCode 和 Claude Code 框架评估；GLM-5.2、Claude Opus 4.8 和 Claude Fable 5 使用 Claude Code 框架；GPT 5.5 和 GPT 5.6 Sol 使用 Codex 框架。所有模型均在最大推理强度下评估，GPT 5.5 除外（使用“xhigh”设置）。

Productivity and agentic benchmarks生产力与智能体基准测试

OfficeQA Pro and SpreadsheetBench 2. Kimi K3, GLM-5.2, Claude Opus 4.8, and Claude Fable 5 are evaluated with the Claude Code harness; GPT 5.5 and GPT 5.6 Sol are evaluated with the Codex harness.OfficeQA Pro 和 SpreadsheetBench 2：Kimi K3、GLM-5.2、Claude Opus 4.8 和 Claude Fable 5 使用 Claude Code 框架评估；GPT 5.5 和 GPT 5.6 Sol 使用 Codex 框架评估。
MCP Atlas. All models are evaluated on the 500-task public subset with a 100-turn limit, using Gemini 3.1 Pro as the judge.MCP Atlas：所有模型均在 500 个任务的公共子集上进行评估，限制为 100 轮，并使用 Gemini 3.1 Pro 作为评判者。
AutomationBench. All models are evaluated on the 600-task public subset, following the official GitHub setup in all other respects.AutomationBench：所有模型均在 600 个任务的公共子集上进行评估，其他方面均遵循 GitHub 官方设置。
BrowseComp. We adopt the context-compaction strategy used in the Claude model cards, triggered at 300K tokens. When evaluated with a 1M-token context window and no context management, Kimi K3 achieves a score of 90.4. The results of Claude Fable 5, Claude Opus 4.8, GPT 5.6 Sol, and GPT 5.5 are cited from https://www.anthropic.com/news/claude-fable-5-mythos-5 and https://openai.com/index/gpt-5-6/.BrowseComp：我们采用 Claude 模型卡中使用的上下文压缩策略，在 300K tokens 时触发。在 1M-token 上下文窗口且无上下文管理的情况下，Kimi K3 获得 90.4 分。Claude Fable 5、Claude Opus 4.8、GPT 5.6 Sol 和 GPT 5.5 的结果引自 https://www.anthropic.com/news/claude-fable-5-mythos-5 和 https://openai.com/index/gpt-5-6/。
GDPval-AA v2 and AA-Briefcase scores are cited from https://artificialanalysis.ai/.GDPval-AA v2 和 AA-Briefcase 分数引自 https://artificialanalysis.ai/。

Multimodal benchmarks多模态基准测试

Except for ZeroBench, which follows the official setting and is run five times, all multimodal scores are averaged over three runs. MMMU-Pro is evaluated following the official protocol, preserving the original input order and prepending images to the text input.除 ZeroBench（遵循官方设置并运行五次）外，所有多模态分数均为三次运行的平均值。MMMU-Pro 评估遵循官方协议，保留原始输入顺序并将图像置于文本输入之前。
PerceptionBench. PerceptionBench is an in-house benchmark that focuses on atomic visual perception capabilities.PerceptionBench：PerceptionBench 是一个内部基准测试，侧重于原子级视觉感知能力。

Limitations局限性

Sensitivity to thinking history. K3 was trained in the preserved thinking history mode. If the agent harness fails to pass back all the historical thinking content as required, or if an ongoing session with another model is switched over to K3, generation quality may become highly unstable. We recommend using a harness with verified compatibility, such as Kimi Code, and avoiding switching to K3 in the middle of a session.对思考历史的敏感性：K3 是在保留思考历史模式下训练的。如果智能体框架未能按要求传回所有历史思考内容，或者在对话中将其他模型切换为 K3，生成质量可能会变得极不稳定。我们建议使用经过验证兼容的框架（如 Kimi Code），并避免在会话中途切换至 K3。
Excessive proactiveness. K3's training places particular emphasis on long-horizon, challenging tasks. As a result, when it encounters minor issues or ambiguous user intent during task execution, it may make unexpected decisions on the user's behalf. If your application requires the agent to operate within well-defined boundaries and refrain from excessive improvisation, please impose more explicit behavioral constraints on K3 in the system prompt or in AGENTS.md.过度主动性：K3 的训练特别强调长程、具有挑战性的任务。因此，当它在任务执行过程中遇到小问题或模糊的用户意图时，可能会代表用户做出意料之外的决策。如果你的应用要求智能体在明确定义的边界内操作且避免过度即兴发挥，请在系统提示词或 AGENTS.md 中对 K3 施加更明确的行为约束。
Despite being a highly competitive model overall, K3 nonetheless exhibits a noticeable gap in user experience compared with Claude Fable 5 and GPT 5.6 Sol.尽管 K3 是一款极具竞争力的模型，但与 Claude Fable 5 和 GPT 5.6 Sol 相比，其用户体验仍存在显著差距。