Skip to content

Welcome to Moonshot AI

icon Moonshot AI is committed to solving ambitious "moonshot" problems that will lead humanity to AGI. We embrace open source, and contributed the following projects to the community:

Research

  • Attention Residuals: a drop-in replacement for residual connections with consistent scaling gains.
  • Kimi K2.5: our most powerful open-source multimodal agentic model designed to advance general agentic intelligence.
  • Kimi Linear: a hybrid linear attention architecture that outperforms traditional full attention methods across various contexts.
  • Kimi K2: an open-source Mixture-of-Experts model with 32B activated parameters and 1T total parameters. It achieves state-of-the-art performance in frontier knowledge, math, and coding among non-thinking models.
  • Mooncake: pioneered the idea of KV-centric disaggregated LLM serving, winning the Best Paper award at FAST 2025. We released code and data corresponding to our paper.
  • Kimi-K1.5: Scaling Reinforcement Learning with LLMs. We released our tech report on building an o1-level multi-modal reasoning model.
  • Moonlight for our paper Muon is Scalable for LLM Training. We released (i) checkpoints for SOTA small models; (ii) code for our improved Muon optimizer.
  • MoBA: Mixture of Block Attention for Long-Context LLMs. We released code and paper for our looong-context LLM technique.
  • Kimi-VL: Mixture-of-Experts Vision-Language Model for Multimodal Reasoning, Long-Context Understanding, and Strong Agent Capabilities
  • Kimina-Prover Preview: Towards Large Formal Reasoning Models with Reinforcement Learning. We released technical report, model weights and a rectified version of miniF2F-test benchmark.
  • Kimi-Audio: Universal audio foundation model that handles diverse tasks like speech recognition, audio understanding, audio-to-text chat, speech-to-speech conversation. We released the technical report, model weights and evaluation toolkit.
  • Kimi-Dev: A Strong and Open-source Coding LLM for Issue Resolution. Kimi-Dev-72B achieves 60.4% performance on SWE-bench Verified.

Agents

  • Kimi Code: A fast, versatile, and extensible AI coding agent that brings Kimi into your projects and development workflow.

Service and Infra

Community

  • Forum: join discussions, ask questions, and share ideas about the Moonshot AI platform and APIs.

Pinned Loading

  1. Kimi-K2.5 Kimi-K2.5 Public

    Moonshot's most powerful model

    2k 253

  2. kimi-code kimi-code Public

    The Starting Point for Next-Gen Agents

    TypeScript 1.7k 168

  3. Kimi-Linear Kimi-Linear Public

    1.4k 72

  4. Kimi-K2 Kimi-K2 Public

    Kimi K2 is the large language model series developed by Moonshot AI team

    10.8k 847

  5. checkpoint-engine checkpoint-engine Public

    Checkpoint-engine is a simple middleware to update model weights in LLM inference engines

    Python 961 84

  6. MoBA MoBA Public

    MoBA: Mixture of Block Attention for Long-Context LLMs

    Python 2.1k 149

Repositories

Showing 10 of 38 repositories
  • kimi-code Public

    The Starting Point for Next-Gen Agents

    MoonshotAI/kimi-code’s past year of commit activity
    TypeScript 1,658 MIT 168 82 26 Updated Jun 4, 2026
  • walle Public
    MoonshotAI/walle’s past year of commit activity
    Go 21 MIT 3 0 0 Updated Jun 1, 2026
  • kimi-cli Public

    Kimi Code CLI is your next CLI agent.

    MoonshotAI/kimi-cli’s past year of commit activity
    Python 8,877 Apache-2.0 1,101 456 291 Updated Jun 1, 2026
  • kimi-agent-sdk Public

    Kimi Agent SDK provides a programmatic interface to interact with the Kimi CLI

    MoonshotAI/kimi-agent-sdk’s past year of commit activity
    TypeScript 481 Apache-2.0 64 66 10 Updated May 29, 2026
  • FlashKDA Public

    FlashKDA: high-performance Kimi Delta Attention kernels

    MoonshotAI/FlashKDA’s past year of commit activity
    Cuda 447 MIT 38 4 1 Updated May 26, 2026
  • .github Public
    MoonshotAI/.github’s past year of commit activity
    2 GPL-3.0 0 0 0 Updated May 26, 2026
  • kimi-help-center Public

    Kimi Help Center is an intelligent Q&A routing system that directs user questions to the right help articles across the full Kimi ecosystem — covering Agent mode, Deep Research, Kimi Code, Kimi Claw, Docs & Sheets, PPT/Slides, membership & billing, and API services.

    MoonshotAI/kimi-help-center’s past year of commit activity
    13 6 1 1 Updated May 8, 2026
  • kosong Public

    The LLM abstraction layer for modern AI agent applications.

    MoonshotAI/kosong’s past year of commit activity
    520 50 7 (1 issue needs help) 0 Updated Apr 28, 2026
  • MoonshotAI/Attention-Residuals’s past year of commit activity
    3,295 182 12 0 Updated Mar 17, 2026
  • MoonshotAI/Branding-Guide’s past year of commit activity
    HTML 2 1 0 0 Updated Mar 11, 2026