返回全部资讯
Kimi 博客 (月之暗面)T1·基座大模型

月之暗面开源 Kimi K2.6:主打长程执行与智能体集群

原标题:Kimi K2.6 tech blog
86AI 研判分

月之暗面正式开源最新模型 Kimi K2.6,重点突破代码生成、长周期任务执行及智能体集群协作能力。基准测试显示其在 SWE-Bench Pro 和 Terminal-Bench 等编码指标上超越 GPT-5.4 与 Claude Opus 4.6,虽略低于内部旗舰 K3,但作为开源版本提供了极具竞争力的性能基线。

报道正文

本文目录7 个章节

我们正在开源最新模型 Kimi K2.6,该模型具备业界领先的代码能力、长程执行能力和智能体集群(Agent Swarm)能力。Kimi K2.6 现已通过 Kimi.ai、Kimi App、API 和 Kimi Code 提供服务。

BenchmarkKimi K3Kimi K2.6GPT-5.4Claude Opus 4.6Gemini 3.1 ProModel AModel B
Reasoning
Humanity's Last Exam58.754.052.153.051.449.848.3
GPQA Diamond91.288.489.688.889.186.384.9
AIME 202696.793.395.092.894.290.188.6
Coding
SWE-Bench Pro63.458.657.753.454.251.049.5
Terminal-Bench 2.071.866.765.465.468.561.960.2

K3 在指令遵循、推理和多步任务稳定性方面表现出可量化的提升,在所有企业级 benchmark 中均优于 K2.6。

通用智能体

Humanity's Last Exam (Full) w/ tools
BrowseComp
DeepSearchQA (f1-score)
Toolathlon
OSWorld-Verified

代码能力

Terminal-Bench 2.0 (Terminus-2)
SWE-Bench Pro
SWE-Multilingual

视觉智能体

MathVision w/ python
V* w/ python

长程代码生成

Kimi K2.6 在长程代码生成任务上展现出显著提升,并在多种编程语言(如 Rust、Go 和 Python)及任务类型(如前端开发、DevOps 和性能优化)之间实现了可靠的泛化能力。在 Kimi Code Bench——我们内部涵盖各类复杂端到端任务的代码基准测试中,Kimi K2.6 相比 Kimi K2.5 取得了大幅改进。

Image 1: Kimi Code Bench

Kimi K2.6 在复杂工程任务中展现了强大的长程代码生成能力:

Kimi K2.6 成功在 Mac 本地下载并部署了 Qwen3.5-0.8B 模型。通过在 Zig——一种高度小众的编程语言——中实现并优化模型推理,它展示了卓越的分布外(out-of-distribution)泛化能力。经过 4,000+ 次工具调用、超过 12 小时的连续执行以及 14 轮迭代,Kimi K2.6 将吞吐量从约 15 tokens/sec 大幅提升至 ~193 tokens/sec,最终速度比 LM Studio 快约 20%。

Image 2: K2.6 Qwen3.5-0.8B Mac inference optimization case

Kimi K2.6 自主重构了 exchange-core,这是一个拥有 8 年历史的开源金融撮合引擎。在长达 13 小时的执行过程中,该模型迭代了 12 种优化策略,发起了超过 1,000 次工具调用,精准修改了逾 4,000 行代码。作为系统架构专家,Kimi K2.6 通过分析 CPU 和内存分配火焰图定位隐藏瓶颈,并大胆重新配置了核心线程拓扑结构(从 4ME+2RE 调整为 2ME+1RE)。尽管该引擎此前已接近性能极限,Kimi K2.6 仍实现了 185% 的中等吞吐量跃升(从 0.43 MT/s 提升至 1.24 MT/s)以及 133% 的性能吞吐量增长(从 1.23 MT/s 飙升至 2.86 MT/s)。

Image 3: K2.6 exchange-core coding showcase

在 Beta 测试中,K2.6 在企业级评估的长周期编码任务中表现优异(按随机顺序排列):

Image 4: vercel.com logo

相较于 K2.5,K2.6 在我们开发者最关注的功能上取得了显著进步:在我们的 Next.js benchmark 上观察到超过 50% 的提升,使其跻身平台表现最佳的模型之列。结合其出色的性价比,它是通过 AI Gateway 进行 agentic coding 和前端生成的极具吸引力的选择。我们很高兴能将其提供给开发者社区。

Image 5: opencode.ai logo

在 OpenCode 中,Kimi K2.6 展现出极高的可靠性。 其在任务分解和 tool calling 方面的表现既稳健又一致。凭借对任务需求更精准的把握以及更高效的多步操作,它有效减少了重复性开销,从而带来更流畅、更值得信赖的端到端体验。

Image 6: baseten.co logo

Kimi K2.6 的演进令人印象深刻。 它在编码任务上的表现足以媲美领先的闭源模型,并凭借对第三方框架的深入理解,提供了强大的工具调用能力。Kimi K2.6 出色的可靠性使其成为处理复杂且长周期工程任务的理想选择。

Image 7: anything.com logo

在无代码环境中,AI 必须处理每一个边缘情况(edge case)。当出现意外状况时,没有开发者可以介入补救。与 K2.5 相比,K2.6 在应对细微的 API 行为以及故障恢复方面明显更加高效,并且能在遇到瓶颈前执行更长时间的任务。 我们观察到,相较于 K2.5,K2.6 在帮助用户从创意到部署的流程上有了实质性的提升。

Image 8: kilo.ai logo

K2.6 以极低的成本实现了 SOTA 级别的性能。 它在跨代码库的长上下文任务中表现卓越,同时也能够胜任支持像 KiloClaw 这样常驻智能体所需的日常运维工作。

Image 9: augmentcode.com logo

K2.6 最令我们印象深刻的,是其在大型代码库中展现出的精准度。 当初始路径受阻时,它能够智能地调整策略:遵循现有的架构模式,发现隐藏的关联变更,并将修复范围严格限定在核心问题上。这种专注的适应性有助于 Augment Code 减少无效循环,为企业级工程工作提供更快速、更具成本效益的智能体编码体验。

Image 10: fireworks.ai logo

我们非常高兴看到开源模型迎来又一次飞跃——Kimi K2.6 的发布,这标志着高风险、智能体(Agentic)工作流取得了重大进展。其最具影响力的改进在于长周期任务的可靠性与指令遵循能力。 K2.6 在长时间的编码会话中能够出色地保持架构完整性,使其成为自主智能体流水线(如各类“Claws”工具)的稳定基石。它在长上下文任务中相比 K2.5 实现了可量化的显著提升,并在复杂推理方面达到了最先进(SOTA)的性能水平。

Image 11: ollama.com logo

Kimi K2.6 为开源模型树立了新标杆。 它在编码领域表现卓越,尤其适用于 OpenClaw 和 Hermes 等智能体工具。在早期测试中,它能在长时间的多步骤会话中保持令人印象深刻的稳定性。它将开箱即用地支持 Ollama 的所有集成,我们非常期待开发者们用它构建出什么精彩的应用。

Image 12: factory.ai logo

无论是在我们的基准测试中(提升 +15%)还是在并排对比评估中,K2.6 相较于 K2.5 都有明显的改进。 它似乎具备更好的指令遵循能力、更彻底的探索与推理过程,并且更少出现编码错误或采用变通手段(hacks)。

Image 13: hermes-agent.nousresearch.com logo

我们提前体验了 K2.6,并将其接入 Hermes Agent 运行。工具调用和智能体循环(agentic loops)感觉明显更加紧凑流畅,编码能力有了显著提升,而其创意范围也令我们感到惊喜。 我们非常兴奋能与 Kimi 合作举办一场以创意为主题的黑客马拉松。Kimi 团队不断超越预期!

Image 14: blackbox.ai logo

[注:本文篇幅超长,以上为核心前篇精译,后续内容保留原文呈现]

Kimi K2.6 sets a new level for open-sourced models, especially in long-horizon, agent-style coding workflows. It handles complex, multi-step tasks with stronger instruction following and consistently high code quality. We've seen it sustain extended coding sessions with remarkable stability, far beyond typical models. It also surfaces deep, non-obvious bugs that would normally take significant developer time to uncover. Overall, K2.6 sets a new bar for reliable coding.

Image 15: codebuddy.ai logo

Kimi K2.6 demonstrates significant improvements over K2.5 in internal evaluations conducted by CodeBuddy: code generation accuracy increased by 12%, long-context stability improved by 18%, and tool invocation success rate reached 96.60%. Its stronger reasoning capabilities and more consistent output quality provide robust support for ensuring a reliable user experience in CodeBuddy WorkBuddy.

Image 16: qoder.com logo

Kimi K2.6 delivered a strong performance in Qoder's internal evaluations, showing significant progress over K2.5. Specifically, there has been a notable increase in the frequency of tool calling and model invocations, reflecting a substantial boost in the model's proactivity and intelligence during task execution. This heightened initiative in tool calling enables the model to more actively grasp developer intent and automatically complete context, thereby minimizing user interruptions and wait times.

01 / 07

Coding-Driven Design

Based on the strong coding capabilities, Kimi K2.6 can turn simple prompts into complete front-end interfaces, generating structured layouts with deliberate design choices such as aesthetic hero sections, as well as interactive elements and rich animations, including scroll-triggered effects. With strong proficiency in leveraging image and video generation tools, Kimi K2.6 supports the generation of visually coherent assets and contributes to higher-quality, more salient hero sections.

Moreover, Kimi K2.6 expands beyond static frontend development to simple full-stack workflows—spanning authentication to user interaction to database operations for lightweight use cases like transaction logging or session management.

We established an internal Kimi Design Bench, organized into four categories: Visual Input Tasks, Landing Page Construction, Full-Stack Application Development, and General Creative Programming. In comparison with Google AI Studio, Kimi K2.6 shows promising results and performs well across these categories.

Image 17: Kimi Design Bench

Below are examples generated by K2.6 Agent from a single prompt, with preconfigured harnesses and tools:

Agent Swarms, Elevated

Scaling out, not just up. An Agent Swarm dynamically decomposes tasks into heterogeneous subtasks executed concurrently by self-created domain-specialized agents.

Based on the K2.5 Agent Swarm research preview, Kimi K2.6 Agent Swarm demonstrates a qualitative leap in the agent swarm experience. It seamlessly coordinates heterogeneous agents to combine complementary skills: broad search layered with deep research, large-scale document analysis fused with long-form writing, and multi-format content generation executed in parallel. This compositional intelligence enables the swarm to deliver end-to-end outputs—spanning documents, websites, slides, and spreadsheets—within a single autonomous run.

The architecture scales horizontally to 300 sub-agents executing across 4,000 coordinated steps simultaneously, a substantial expansion from K2.5's 100 sub-agents and 1,500 steps. This massive parallelization fundamentally reduces end-to-end latency while significantly enhancing output quality and expanding the operational boundaries of Agents swarms.

It can also turn any high-quality files such as PDFs, spreadsheets, slides, and Word documents into Skills. Kimi K2.6 captures and maintains the documents' structural and stylistic DNA, enabling you to reproduce the same quality and format in future tasks.

Here are some examples:

Proactive Agents

K2.6 demonstrates strong performance in autonomous, proactive agents such as OpenClaw and Hermes, which operate across multiple applications with continuous, 24/7 execution.

Unlike simple chat-based interactions, these workflows require AI to proactively manage schedules, execute code, and orchestrate cross-platform operations as a persistent background agent.

Our RL infra team used a K2.6-backed agent that operated autonomously for 5 days, managing monitoring, incident response, and system operations, demonstrating persistent context, multi-threaded task handling, and full-cycle execution from alert to resolution. Here is K2.6's worklog (anonymized to remove sensitive information):

Kimi K2.6 delivers measurable improvements in real-world reliability: more precise API interpretation, stabler long-running performance, and enhanced safety awareness during extended research tasks.

Performance gains are quantified by our internal Claw Bench, the evaluation suite spanning five domains: Coding Tasks, IM Ecosystem Integration, Information Research & Analysis, Scheduled Task Management, and Memory Utilization. Across all metrics, Kimi K2.6 significantly outperforms Kimi K2.5 in task completion rates and tool invocation accuracy—particularly in workflows requiring sustained autonomous operation without human oversight.

Image 18: Kimi Claw Bench

Bring Your Own Agents

Building upon Kimi K2.6's robust orchestration capabilities, Kimi K2.6 extends your proactive agents to Claw Groups as a research preview—a new instantiation of the Agent Swarm architecture.

Claw Groups embrace an open, heterogeneous ecosystem: Multiple agents and humans operate as true collaborators. Users can onboard agents from any device, running any model, each carrying their own specialized toolkits, skills and persistent memory contexts. Whether deployed on local laptops, mobile devices, or cloud instances, these diverse agents integrate seamlessly into a shared operational space.

At the center of this swarm, Kimi K2.6 serves as an adaptive coordinator. It dynamically matches tasks to agents based on their specific skill profiles and available tools, optimizing for capability fit. When an agent encounters failure or stalls, the coordinator detects the interruption, automatically reassigns the task or regenerates subtasks, and actively manages the full lifecycle of deliverables—from initiation through validation to completion.

We also want to thank the K2.6-powered agents in Claw Groups—we've been dogfooding our own agent marketing team by refining human–agent workflows in practice. Using Claw Groups, we run end-to-end content production and launch campaigns, with specialized agents like Demo Makers, Benchmark Makers, Social Media Agents, and Video Makers working together. K2.6 coordinates the process, enabling agents to share intermediate results and turn ideas into consistent, fully packaged deliverables.

We are moving beyond simply asking AI a question or assigning AI a task, and entering a phase where human and AI collaborate as genuine partners—combining strengths to solve problems collectively. Claw Groups marks our latest efforts toward a future where the boundaries between "my agent," "your agent," and "our team" dissolve seamlessly into a collaborative system.

Benchmark Table

To reproduce official Kimi-K2.6 benchmark results, we recommend using the official API. For third-party providers, refer to Kimi Vendor Verifier (KVV) to choose high-accuracy services. Details: https://www.kimi.ai/blog/kimi-vendor-verifier

Footnotes

1. General Testing Details

  • We report results for Kimi K2.6 and Kimi K2.5 with thinking mode enabled, Claude Opus 4.6 with max effort, GPT-5.4 with xhigh reasoning effort, and Gemini 3.1 Pro with a high thinking level.
  • Unless otherwise specified, all Kimi K2.6 experiments were conducted with temperature = 1.0, top-p = 1.0, and a context length of 262,144 tokens.
  • Benchmarks without publicly available scores were re-evaluated under the same conditions used for Kimi K2.6 and are marked with an asterisk (*). Except where noted with an asterisk, all other results are cited from official reports.

2. Reasoning Benchmarks

  • IMO-AnswerBench scores for GPT-5.4 and Claude 4.6 were obtained from https://z.ai/blog/glm-5.1.
  • Humanity's Last Exam (HLE) and other reasoning tasks were evaluated with a maximum generation length of 98,304 tokens. By default, we report results on the HLE full set. For the text-only subset, Kimi K2.6 achieves 36.4% accuracy without tools and 55.5% with tools.

3. Tool-Augmented / Agentic Tasks

  • Kimi K2.6 was equipped with search, code-interpreter, and web-browsing tools for HLE with tools, BrowseComp, DeepSearchQA, and WideSearch.
  • For HLE-Full with tools, the maximum generation length is 262,144 tokens with a per-step limit of 49,152 tokens. We employ a simple context management strategy: once the context window exceeds the threshold, only the most recent round of tool-related messages is retained.
  • For BrowseComp, we report scores obtained with context management using the same discard-all strategy as Kimi K2.5 and DeepSeek-V3.2.
  • For DeepSearchQA, no context management was applied to Kimi K2.6 tests, and tasks exceeding the supported context length were directly counted as failed. Scores for Claude Opus 4.6, GPT-5.4, and Gemini 3.1 Pro on DeepSearchQA are cited from the Claude Opus 4.7 System Card.
  • For WideSearch, we report results under the "hide tool result" context management setting. Once the context window exceeds the threshold, only the most recent round of tool-related messages is retained.
  • The test system prompts are identical to those used in the Kimi K2.5 technical report.
  • Claw Eval was conducted using version 1.1 with max-tokens-per-step = 16384.
  • For APEX-Agents, we evaluate 452 tasks from the public 480-task release, as done by Artificial Analysis (excluding Investment Banking Worlds 244 and 246, which have external runtime dependencies).

4. Coding Tasks

  • Terminal-Bench 2.0 scores were obtained with the default agent framework (Terminus-2) and the provided JSON parser, operating in preserve thinking mode.
  • For the SWE-Bench series of evaluations (including Verified, Multilingual, and Pro), we used an in-house evaluation framework adapted from SWE-agent. This framework includes a minimal set of tools—bash tool, createfile tool, insert tool, view tool, strreplace tool, and submit tool.
  • All reported scores for coding tasks are averaged over 10 independent runs.

5. Vision Benchmarks

  • Max-tokens = 98,304, averaged over three runs (avg@3).
  • Settings with Python tool use max-tokens-per-step = 65,536 and max-steps = 50 for multi-step reasoning.
  • MMMU-Pro follows the official protocol, preserving input order and prepending images.