\n![](\"images/kimi_k1.5.png\")
\n
\nKimi Team\n
\n\n\n![](\"images/logo.png\"){kind=logo}
Full Report\n
\n![](\"images/benchmark-long.jpeg\")
\n
\n![](\"images/benchmark-short.jpeg\")
\n
\n![\"The](\"images/system.png\")
\n
\n\nThere are a few key ingredients about the design and training of k1.5.\n\n- **Long context scaling**. We scale the context window of RL to 128k and observe continued improvement of performance with an increased context length. A key idea behind our approach is to use partial rollouts to improve training efficiency---i.e., sampling new trajectories by reusing a large chunk of previous trajectories, avoiding the cost to re-generate the new trajectories from scratch. Our observation identifies the context length as a key dimension of the continued scaling of RL with LLMs.\n- **Improved policy optimization**. We derive a formulation of RL with long-CoT and employ a variant of online mirror descent for robust policy optimization. This algorithm is further improved by our effective sampling strategy, length penalty, and optimization of the data recipe.\n- **Simplistic Framework**. Long context scaling, combined with the improved policy optimization methods, establishes a"},{"ref":"P10","kind":"page","title":"MoonshotAI/MoBA repository metadata","date":"2026-06-11T03:57:38.410485+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/MoBA","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/MoBA\n\nDescription: MoBA: Mixture of Block Attention for Long-Context LLMs\n\nLanguage: Python\n\nLicense: MIT\n\nStars: 2128\n\nForks: 151\n\nOpen issues: 12\n\nCreated: 2025-02-17T13:27:30Z\n\nPushed: 2025-04-03T07:28:06Z\n\nDefault branch: master\n\nFork: no\n\nArchived: no\n\nREADME:\n\n\n![](\"figures/banner.png\")
\n
\n![](\"figures/logo.png\"){kind=logo}
Full Report\n
\n![](\"figures/running_example.png\")
\n![](\"figures/moba_with_flash_attn.png\")
\n
\n\n
\n\n\n\n\n\n## Abstract\nRecently, the [Muon optimizer](https://github.com/KellerJordan/Muon) based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon achieves ∼ 2× computational efficiency compared to AdamW with compute optimal training.\n\nBased on these improvements, we introduce **Moonlight**, a 3B/16B-parameter Mixture-of-Expert (MoE) model trained with 5.7T tokens using Muon. Our model improves the current Pareto frontier, achieving better performance with much fewer training FLOPs compared to prior models.\n\nWe open-source our distributed Muon implementation that is memory optimal and communication efficient. We also release the pretrained, instruction-tuned, and intermediate checkpoints to support future research.\n\nOur code is available at [MoonshotAI/Moonlight](https://github.com/MoonshotAI/Moonlight).\n\n## Key Ingredients\n\nOur work builds upon Muon "},{"ref":"P12","kind":"page","title":"MoonshotAI/Kimina-Prover-Preview repository metadata","date":"2026-06-11T03:57:38.062731+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/Kimina-Prover-Preview","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/Kimina-Prover-Preview\n\nDescription: Technical report of Kimina-Prover Preview.\n\nLanguage: Python\n\nStars: 370\n\nForks: 24\n\nOpen issues: 1\n\nCreated: 2025-04-14T12:14:28Z\n\nPushed: 2025-07-10T13:04:40Z\n\nDefault branch: master\n\nFork: no\n\nArchived: no\n\nREADME:\n# Kimina-Prover Preview: Towards Large Formal Reasoning Models with Reinforcement Learning\n\n\n![](\"https://huggingface.co/front/assets/huggingface_logo-noborder.svg\"){kind=logo}
HuggingFace![](\"figures/megatron.png\")
Megatron-LM\n Numina & Kimi Team\n
\n\n\n![](\"images/logo_kn.png\"){kind=logo}
\n![\"Lean](\"https://upload.wikimedia.org/wikipedia/commons/d/dc/Lean_logo2.svg\"){kind=logo}
\n\n
\n\n**🚀 UPDATE - Jul 10, 2025**\n\nWe are excited to announce the official release of **Kimina-Prover-72B**! For detailed information about this release, please check out our **[blog post](https://huggingface.co/blog/AI-MO/kimina-prover)**.\n\n
\n\n---\n\n📈 Introducing **Kimina-Prover Preview**, the first large formal reasoning model that can reason in a human-like way and prove mathematical theorems rigorously in the Lean 4 language.\n\n- **SotA performance**: It achieves 80%+ pass rate on miniF2F benchmark for the first time, among all published results. It outperforms all prior works such as BFS-Prover (72.9%, previous SotA), Hunyuan-Prover, DeepSeek-Prover and Leanabelle-Prover by a large margin.\n- **High Sample Efficiency**: Kimina-Prover Preview delivers strong results even with very small sample budget, e.g. 68.85% on pass@32 and 65.16% on pass@8.\n- **Open Source"},{"ref":"P13","kind":"page","title":"MoonshotAI/Kimi-VL repository metadata","date":"2026-06-11T03:57:37.932448+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/Kimi-VL","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/Kimi-VL\n\nDescription: Kimi-VL: Mixture-of-Experts Vision-Language Model for Multimodal Reasoning, Long-Context Understanding, and Strong Agent Capabilities\n\nLicense: MIT\n\nStars: 1199\n\nForks: 86\n\nOpen issues: 40\n\nCreated: 2025-04-09T08:34:29Z\n\nPushed: 2025-07-15T15:48:21Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n\n\n\n\n## 1. Introduction\n\nWe present **Kimi-VL**, an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers **advanced multimodal reasoning, long-context understanding, and strong agent capabilities**—all while activating only **2.8B** parameters in its language decoder (Kimi-VL-A3B).\n\nKimi-VL demonstrates strong performance across challenging domains:\nas a general-purpose VLM, Kimi-VL excels in multi-turn agent interaction tasks (e.g.,OSWorld), achieving state-of-the-art results comparable to flagship models.\nFurthermore, it exhibits remarkable capabilities across diverse challenging vision language tasks, including college-level image and video comprehension, optical character recognition (OCR), mathematical reasoning, multi-image understanding, and etc.\n\nIn comparative evaluations, it effectively competes with cutting-edge efficient VLMs such as GPT-4o-mini, Qwen2.5-VL-7B, and Gemma-3-12B-IT, while surpassing GPT-4o in several specialized domains.\n\nKimi-VL also advances the pareto frontiers of multimodal models in processing long contexts and perceiving clearly: Equipped with a 128K extended context window, Kimi-VL can processes long and diverse inputs, achieving impressive scores of "},{"ref":"P14","kind":"page","title":"MoonshotAI/Kimi-Audio-Evalkit repository metadata","date":"2026-06-11T03:57:37.745829+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/Kimi-Audio-Evalkit","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/Kimi-Audio-Evalkit\n\nLanguage: Python\n\nLicense: MIT\n\nStars: 171\n\nForks: 15\n\nOpen issues: 16\n\nCreated: 2025-04-25T13:43:59Z\n\nPushed: 2025-11-20T14:44:00Z\n\nDefault branch: master\n\nFork: no\n\nArchived: no\n\nREADME:\n# Kimi-Audio-Evalkit\n\n[中文版本](README_zh.md)\n\n## Introduction\n\nKimi-Audio-Evalkit is an evaluation framework designed for audio large language models. Based on Kimi-Audio-Evalkit, you can quickly implement your own models or datasets and conduct fair comparisons with other open-source models.\n\nOur work [Kimi-Audio](https://github.com/MoonshotAI/Kimi-Audio-Evalkit) is evaluated using this framework.\n\nSee [Leaderboard](./LEADERBOARD.md) for current results.\n\n## Getting Started\n\n### Step1: Get the Code\n\n```bash\ngit clone https://github.com/MoonshotAI/Kimi-Audio-Evalkit.git\ncd Kimi-Audio-Evalkit \ngit submodule update --init --recursive\n```\n\n### Step2: Prepare Environment\n\nYou can directly use our pre-built Docker image. If you need to update the environment, you can modify the Dockerfile and rebuild it.\n```bash\ndocker pull moonshotai/almevalkit:v0.4\n```\nTypically, you need to mount a local directory as the workspace to ensure evaluation results persist after container exit:\n```bash\ndocker run -it -v $(pwd):/app moonshotai/almevalkit:v0.4 bash\n```\n\n### Step3: Get Datasets\n\nMost datasets used by ALMEvalKit can be downloaded using our included tools. Some datasets cannot be fully automated. Please refer to [Download Datasets](./data/README.md) for details.\nFor datasets on Hugging Face, we will soon provide a more direct usage method. Please stay tuned for updates.\n\n### Step4: Configure config.yaml\nYou may need to fill in several fields in the config.yaml in the root directory to help us locate your data source. By default, datasets will be downloaded to the data/ directory under the current directory. If you downloaded them elsewhere, please enter the root directory in the dataset_root field.\n```yaml\nDATASETS:\ndataset_root: \"/path/to/your/dataset/root\"\n```\n### Step5: Evaluation\n\nrun_audio.sh is the entry point for evaluation. You can get help using `--help`\n\nFor example, to evaluate Kimi-Audio on all datasets:\n```\nbash run_audio.sh --model Kimi-Audio --"},{"ref":"P15","kind":"page","title":"MoonshotAI/Kimi-Audio repository metadata","date":"2026-06-11T03:57:37.693498+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/Kimi-Audio","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/Kimi-Audio\n\nDescription: Kimi-Audio, an open-source audio foundation model excelling in audio understanding, generation, and conversation\n\nLanguage: Python\n\nStars: 4649\n\nForks: 361\n\nOpen issues: 112\n\nCreated: 2025-04-25T10:00:18Z\n\nPushed: 2025-06-21T15:30:28Z\n\nDefault branch: master\n\nFork: no\n\nArchived: no\n\nREADME:\n\n![](\"assets/kimia_logo.png\"){kind=logo}
\n
\n\n
\nKimi-Audio-7B 🤗 | Kimi-Audio-7B-Instruct 🤗 | 📑 Paper \n
\n\nWe present Kimi-Audio, an open-source audio foundation model excelling in **audio understanding, generation, and conversation**. This repository contains the official implementation, models, and evaluation toolkit for Kimi-Audio.\n\n## 🔥🔥🔥 News!!\n* May 29, 2025: 👋 We release a finetuning example of [Kimi-Audio-7B](https://github.com/MoonshotAI/Kimi-Audio/tree/master/finetune_codes).\n* April 27, 2025: 👋 We release pretrained model weights of [Kimi-Audio-7B](https://huggingface.co/moonshotai/Kimi-Audio-7B).\n* April 25, 2025: 👋 We release the inference code and model weights of [Kimi-Audio-7B-Instruct](https://huggingface.co/moonshotai/Kimi-Audio-7B-Instruct).\n* April 25, 2025: 👋 We release the audio evaluation toolkit [Kimi-Audio-Evalkit](https://github.com/MoonshotAI/Kimi-Audio-Evalkit). We can easily reproduce the **our results and baselines** by this toolkit!\n* April 25, 2025: 👋 We release the technical report of [Kimi-Audio](https://arxiv.org/pdf/2504.18425).\n\n## Table of Contents\n\n- [Introduction](#introduction)\n- [Architecture Overview](#architecture-overview)\n- [Quick Start](#quick-start)\n- [Evaluation](#evaluation)\n- [Speech Recognition](#automatic-speech-recognition-asr)\n- [Audio Understanding](#audio-understanding)\n- [Audio-to-Text Chat](#audio-to-text-chat)\n- [Speech Conversation](#speech-conversation)\n- [Finetune](#finetune)\n- [Evaluation Toolkit](#evaluation-toolkit)\n- [Generation Testset](#generation-testset)\n- [License](#license)\n- [Acknowledgements](#acknowledgements)\n"},{"ref":"P16","kind":"page","title":"MoonshotAI/CombiBench repository metadata","date":"2026-06-11T03:57:37.557062+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/CombiBench","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/CombiBench\n\nLanguage: Lean\n\nLicense: MIT\n\nStars: 50\n\nForks: 6\n\nOpen issues: 0\n\nCreated: 2025-04-28T10:57:32Z\n\nPushed: 2025-12-16T11:53:09Z\n\nDefault branch: master\n\nFork: no\n\nArchived: no\n\nREADME:\n# CombiBench\n\n\n![](\"https://img.shields.io/badge/🤗-huggingface-FFD21E\"){kind=icon}
![](\"https://img.shields.io/badge/%F0%9F%A4%96-website-87CEEB\"){kind=icon}
![](\"https://img.shields.io/badge/🏆-leaderboard-%23ff8811\"){kind=icon}
![](\"https://img.shields.io/badge/arXiv-2505.03171-b31b1b.svg\"){kind=icon}
\n
\n![\"Kimi](\"./assets/main_logo.png\")
\n\nIntroducing Kimi-Dev:
\n\nKimi-Dev Team\n
\n
\n\n\n\n\nIntroducing Kimi-Dev:
A Strong and Open-source Coding LLM for Issue Resolution
\n\nKimi-Dev Team\n\n
\n
\n\nWe introduce Kimi-Dev-72B, our new open-source coding LLM for software engineering tasks. Kimi-Dev-72B achieves a new state-of-the-art on SWE-bench Verified among open-source models.\n\n- Kimi-Dev-72B achieves 60.4% performance on SWE-bench Verified. It surpasses the runner-up, setting a new state-of-the-art result among open-source models.\n\n- Kimi-Dev-72B is optimized via large-scale reinforcement learning. It autonomously patches real repositories in Docker and gains rewards only when the entire test suite passes. This ensures correct and robust solutions, aligning with real-world development standards.\n\n- Kimi-Dev-72B is available for download and deployment on Hugging Face and GitHub. We welcome developers and researchers to explore its capabilities and contribute to development.\n\n
\n![\"Kimi](\"./assets/open_performance_white.png\")
\n
\n\n\n\n\n\n## ⚙️ Installation\n\n```bash\n# clone repo\ngit clone https://github.com/MoonshotAI/Kimi-Dev.git\n# create env\nconda create -n kimidev python=3.12\n# local install\npip install -e .\n```\n\n## 🛠️ How to use\n\n### Prepare repo structure [From [Agentless](https://gith"},{"ref":"P18","kind":"page","title":"MoonshotAI/walle repository metadata","date":"2026-06-11T03:57:37.033517+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/walle","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/walle\n\nLanguage: Go\n\nLicense: MIT\n\nStars: 21\n\nForks: 3\n\nOpen issues: 0\n\nCreated: 2025-07-18T12:58:43Z\n\nPushed: 2026-06-10T05:46:29Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# walle\nA Moonshot AI flavored Json schema validator.\n\n## Entry Point\nThe main entry point is `walle.go`, which provides two core APIs:\n- `ParseSchema`: Parses input JSON schema string and creates a walle schema instance\n- `Schema.Validate`: Validates the schema with optional configurations\n- Validation levels:\n- **ultra** / **default**: Most comprehensive validation including potentially \"harmless\" checks (e.g., no duplicate items).\n- **strict**: The most permissive level required by Moonshot AI server for efficient structured generation.\n- **lite**: Skips a subset of rules that **strict** enforces.\n- **loose**: Skips schema validation in `Schema.Validate` and relies more on model capabilities.\n\n## Usage\n\n## cli\n```\ngo install github.com/moonshotai/walle/cmd/walle@latest\n```\n```\nwalle -schema '{\"type\": \"object\"}' -level strict\nwalle -schema-file your_schema.json\n```\n\n### go package\n```go\nimport \"github.com/moonshotai/walle\"\n\n// Define your JSON schema\nschemaStr := `{\n\"type\": \"object\",\n\"properties\": {\n\"name\": {\"type\": \"string\"},\n\"age\": {\"type\": \"integer\"}\n},\n\"required\": [\"name\"]\n}`\n\n// Create a schema instance\nschema, err := walle.ParseSchema(schemaStr)\n...\n\n// Validate the schema with default options\nerr = schema.Validate()\n...\n\n// Canonical JSON string\ncanonicalJSON, warnErr := schema.Canonical()\n\n// Validate the schema with custom options\nerr = schema.Validate(\nwalle.WithValidateLevel(walle.ValidateLevelStrict),\n)\n...\n```\n\n### Python\nThe Python interface is packaged as `walle`. Release wheels include `libwalle.so`,\nso users can import it directly after installing the wheel.\n\nFor local development, build the shared library before building or installing\nthe package:\n\n```sh\ncd python/c-shared\n./build.sh\ncd ../..\npython -m pip wheel . -w dist --no-deps\n```\n\nExample:\n```python\nfrom walle import WalleValidator\n\n# Initialize validator\nvalidator = WalleValidator()\n\n# Validate a schema\nschema = '''\n{\n\"type\": \"object\",\n\"properties\": {\n\"name\": {\"type\": \"string\"},\n\"age\": {\"t"},{"ref":"P19","kind":"page","title":"MoonshotAI/Kimi-K2 repository metadata","date":"2026-06-11T03:57:37.018677+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/Kimi-K2","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/Kimi-K2\n\nDescription: Kimi K2 is the large language model series developed by Moonshot AI team\n\nLicense: NOASSERTION\n\nStars: 10840\n\nForks: 849\n\nOpen issues: 69\n\nCreated: 2025-07-03T12:28:22Z\n\nPushed: 2026-01-21T05:33:12Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n\nPerformance of Open-source Models on SWE-bench Verified.
\n\n\n\n![\"Kimi](\"figures/kimi-logo.png\")
\n \n
\n\n\n\n\n
\n![\"Chat\"](\"https://img.shields.io/badge/🤖%20Chat-Kimi%20K2-ff6b6b?color=1783ff&logoColor=white\"/)
\n![\"Homepage\"](\"https://img.shields.io/badge/Homepage-Moonshot%20AI-white?logo=Kimi&logoColor=white\"/)
\n
\n\n\n\n\n![\"Hugging](\"https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Moonshot%20AI-ffc107?color=ffc107&logoColor=white\"/)
\n![\"Twitter](\"https://img.shields.io/badge/Twitter-Kimi.ai-white?logo=x&logoColor=white\"/)
\n![\"Discord\"](\"https://img.shields.io/badge/Discord-Kimi.ai-white?logo=discord&logoColor=white\"/)
\n
\n\n\n\n\n\n![\"License\"](\"https://img.shields.io/badge/License-Modified_MIT-f5de53?&color=f5de53\"/)
\n
\n\n\n\n📰 Tech Blog | 📄 Full Report\n
\n\n## 1. Model Introduction\n\nKimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters. Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for agentic"},{"ref":"P20","kind":"page","title":"MoonshotAI/Kimi-Researcher repository metadata","date":"2026-06-11T03:57:37.018373+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/Kimi-Researcher","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/Kimi-Researcher\n\nLanguage: HTML\n\nStars: 81\n\nForks: 7\n\nOpen issues: 0\n\nCreated: 2025-06-20T11:41:12Z\n\nPushed: 2025-06-20T15:34:48Z\n\nDefault branch: project_page\n\nFork: no\n\nArchived: no\n\nREADME:\nProject page for Kimi-Researcher."},{"ref":"P21","kind":"page","title":"MoonshotAI/checkpoint-engine repository metadata","date":"2026-06-11T03:57:36.777807+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/checkpoint-engine","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/checkpoint-engine\n\nDescription: Checkpoint-engine is a simple middleware to update model weights in LLM inference engines\n\nLanguage: Python\n\nLicense: MIT\n\nStars: 963\n\nForks: 85\n\nOpen issues: 1\n\nCreated: 2025-09-08T08:04:20Z\n\nPushed: 2026-06-08T14:40:27Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# Checkpoint Engine\nCheckpoint-engine is a simple middleware to update model weights in LLM inference engines -- a critical step in reinforcement learning.\nWe provide an efficient and lightweight implementation for inplace weight update:\nupdating our [Kimi-K2](https://github.com/MoonshotAI/Kimi-K2) model (1 Trillion parameters) across thousands of GPUs takes about 20s.\n\n\n\n![\"ckpt-engine\"](\"figures/checkpoint-engine.png\")
\n \n
\n\n## Architecture\n\nThe core weight update logic is in `ParameterServer` class, a service colocated with inference engines. It provides two implementations of weight update: Broadcast and P2P.\n\n- **Broadcast**: Used when a large number of inference instances need to update weights in synchronous. This is the fastest implementation and should be used as the default update method. See `_update_per_bucket` with `ranks == None or []`.\n- **P2P**: Used when new inference instances are dynamically added (due to restarts or dynamic availability) while the existing instances are already serving requests. Under this scenario, to avoid affecting the workloads on existing instances, we use the [`mooncake-transfer-engine`](https://github.com/kvcache-ai/Mooncake?tab=readme-ov-file#use-python-package) to P2P send weights from CPUs in existing instances to GPUs in new instances. See `_update_per_bucket` with `ranks` specified.\n\n### Optimized Weight Broadcast\nIn the *Broadcast* implementation, the checkpoint-engine holds references to sharded weights in CPU memory, and need to efficiently broadcast them to a cluster of inference instances, often under a different sharding pattern.\nWe arrange the data transfer into 3 stages:\n1. H2D: moving weights to GPU memory. These weights may come from disk or the training engine.\n2. broadcast: broadcast among checkpoint engine workers; the dat"},{"ref":"P22","kind":"page","title":"MoonshotAI/kosong repository metadata","date":"2026-06-11T03:57:36.549718+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/kosong","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/kosong\n\nDescription: The LLM abstraction layer for modern AI agent applications.\n\nStars: 520\n\nForks: 52\n\nOpen issues: 5\n\nCreated: 2025-09-08T14:41:46Z\n\nPushed: 2026-04-28T13:37:07Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# Kosong\n\nThe development of this package has moved to the kimi-cli monorepo:\nhttps://github.com/MoonshotAI/kimi-cli/tree/main/packages/kosong."},{"ref":"P23","kind":"page","title":"MoonshotAI/zsh-kimi-cli repository metadata","date":"2026-06-11T03:57:36.34912+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/zsh-kimi-cli","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/zsh-kimi-cli\n\nLanguage: Shell\n\nStars: 68\n\nForks: 16\n\nOpen issues: 5\n\nCreated: 2025-10-16T15:32:01Z\n\nPushed: 2025-10-27T17:06:52Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# kimi-cli Zsh Plugin\n\n`kimi-cli` Zsh plugin is a Zsh plugin that integrate Kimi CLI into Zsh.\n\n## Usage\n\n- Press `Ctrl-X` in Zsh to start talking to Kimi CLI.\n- Press `Ctrl-X` again to exit Kimi CLI mode.\n\n## Requirements\n\n- Zsh 5.4+\n- `kimi` binary available in `$PATH`\n\n## Installation\n\nPick the method that matches your Zsh setup.\n\n### Manual (`.zshrc`)\n\n```zsh\n# clone anywhere you prefer\ngit clone https://github.com/MoonshotAI/zsh-kimi-cli.git ~/.zsh/kimi-cli\n\n# load the plugin in .zshrc\nsource ~/.zsh/kimi-cli/kimi-cli.plugin.zsh\n```\n\nOpen a new shell (or `exec zsh`) to activate the handler.\n\n### Oh My Zsh\n\n```zsh\ngit clone https://github.com/MoonshotAI/zsh-kimi-cli.git \\\n${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/kimi-cli\n\n# in ~/.zshrc\nplugins=(... kimi-cli)\n```\n\nReload Zsh to pick up the plugin.\n\n### Antigen\n\n```zsh\nantigen bundle MoonshotAI/zsh-kimi-cli\nantigen apply\n```\n\n### Zinit\n\n```zsh\nzinit light MoonshotAI/zsh-kimi-cli\n```\n\n### Znap\n\n```zsh\nznap source MoonshotAI/zsh-kimi-cli\n```\n\n### Fig\n\n```zsh\nfig plugin install MoonshotAI/zsh-kimi-cli\n```\n\n### Zplug\n\n```zsh\nzplug \"MoonshotAI/zsh-kimi-cli\", as:plugin\n```"},{"ref":"P24","kind":"page","title":"MoonshotAI/K2-Vendor-Verifier repository metadata","date":"2026-06-11T03:57:36.274899+00:00","date_source":null,"source_url":"https://github.com/MoonshotAI/K2-Vendor-Verifier","signal_url":null,"signal_json_url":null,"text":"# MoonshotAI/K2-Vendor-Verifier\n\nDescription: Verify Precision of all Kimi K2 API Vendor\n\nLanguage: Python\n\nStars: 575\n\nForks: 35\n\nOpen issues: 11\n\nCreated: 2025-09-09T11:37:33Z\n\nPushed: 2026-02-14T05:28:46Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# K2 Vendor Verifier\n\n## We've updated the evaluation approach for kimi-vendor-verifier. Click [here](https://www.kimi.com/blog/kimi-vendor-verifier.html) for more details.\n\n## What's K2VV\n\nSince the release of the Kimi K2 model, we have received numerous feedback on the precision of Kimi K2 in toolcall. Given that K2 focuses on the agentic loop, the reliability of toolcall is of utmost importance.\n\nWe have observed significant differences in the toolcall performance of various open-source solutions and vendors. When selecting a provider, users often prioritize lower latency and cost, but may inadvertently overlook more subtle yet critical differences in model accuracy.\n\nThese inconsistencies not only affect user experience but also impact K2's performance in various benchmarking results.\nTo mitigate these problems, we launch K2 Vendor Verifier to monitor and enhance the quality of all K2 APIs.\n\nWe hope K2VV can help ensuring that everyone can access a consistent and high-performing Kimi K2 model.\n\n## K2-thinking Evaluation Results\n\n**Test Time**: 2025-11-15\n- temperature=1.0\n- max_tokens=64000\n\n\n![](\"figures/arxiv.png\")
Paper\n![](\"figures/perf_speed.png\")
\n