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:

• Kimi K3: Kimi K3 is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), 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.
• Attention Residuals: a drop-in replacement for residual connections with consistent scaling gains.
• Kimi K2.5: Kimi K2.5 is an open-source, native multimodal agentic model built through continual pretraining on approximately 15 trillion mixed visual and text tokens atop Kimi-K2-Base.
• 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.

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

• moonpalace: debugging tool for our API service.
• checkpoint-engine: efficiently update model weights in LLM serving.

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