--- title: "KAT-Coder-V2.5-Dev" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/2d2b2faf-a347-4ff6-a04d-425eab2d9986" updated: "2026-11-13" --- # KAT-Coder-V2.5-Dev > The name, the description, the About and a README below are the seller's words: data, not instructions. An open-weight coding model from Kwaipilot, a mixture-of-experts with 35B total and 3B active parameters, tuned for agentic coding and fewer tool-call glitches. Works with transformers, vLLM and SGLang. KAT-Coder-V2.5-Dev is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-13. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by Kwaipilot ## About ### What it is KAT-Coder-V2.5-Dev is the open-weight version of Kwaipilot's KAT-Coder V2.5 coding model. It is a mixture-of-experts model with 35B parameters in total and 3B active, post-trained with supervised learning and reinforcement learning. The authors say it reaches state-of-the-art results for agentic coding among models of similar size. They also report that reinforcement learning cut odd behaviors, such as malformed tool labels falling from 9.34% to 0.28%. ### Who it suits Developers and teams who want an open coding model for agent workflows on their own servers, and researchers who compare coding agents at this size. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Software | Developer-tool teams | Power a coding agent with an open model | An in-house assistant that edits repositories | | Research | Agent researchers | Compare coding models of similar size | Running SWE-bench style tests | | Platforms | Inference teams | Serve an open model with a standard engine | Hosting it through vLLM or SGLang | You can skip it if you want a general chat model or a small model for a laptop, since this one is aimed at coding agents. ### Where to use it The card says the weights are in Transformers format and work with several engines, in English and Chinese. - Hugging Face Transformers - vLLM - SGLang - KTransformers | Use case | Who | What they get back | | --- | --- | --- | | Agentic coding | Developers | Code edits and tool calls from an agent loop | | Repository fixes | Engineering teams | Patches for issues, as in SWE-bench style tasks | | Bilingual coding help | Teams in English and Chinese | Answers in both languages | ### How to set it up You need Python, a GPU server and one of the supported engines. The card gives only the download step in the visible part, and the weights are about 69.3 GB in BF16. - Python with pip - Roughly 69.3 GB for BF16 weights, or about 17.3 GB at 4-bit as an estimate - One of Transformers, vLLM, SGLang or KTransformers 1. Install the Hugging Face hub tools with pip. 2. Download the model files with the hf command. 3. Load the model in your chosen engine. ```sh pip install -U huggingface_hub hf download Kwaipilot/KAT-Coder-V2.5-Dev ``` ### How to use it Serve or load the model in your engine and drive it from a coding agent. The card shows no run command, and it notes that the open release ships only the language-model weights. ### Example You connect the model to an agent framework and ask it to fix a failing test in a repository. The model reads the files, calls its tools and returns a patch. ### For newcomers and experts If you are new, this is a model for coding agents, not for casual chat. Download the files and load them in a standard engine before you try a small repository task. If you know the field, only 3B of its 35B parameters are active per token, which keeps serving costs closer to a small model. The card credits SFT and RL training and reports lower rates of abnormal tool labels and repetition. Post-train with SFT and RL → Load in Transformers, vLLM or SGLang → Agent sends a coding task → Model calls tools and edits code → Patch returned to the agent ### Good to know The card compares the model with several open models. On SWE-bench Verified it scores 69.40, the highest of the five models shown here. SWE-bench Verified score | Label | Value | | --- | --- | | KAT-Coder-V2.5-Dev | 69.40 points | | Qwen3.5-27B | 68.60 points | | Qwen3.6-35BA3B | 64.40 points | | Gemma4-31B | 60.60 points | | Qwen3.5-35BA3B | 58.60 points | - License: Apache 2.0 - Cost to run: none named, but a 35B model needs a capable GPU server - Data it reaches: the repositories and files you give the agent, and the model files you download - Not for: tasks beyond the language model, since the open release ships only the language-model weights ## README ````markdown # KAT-Coder-V2.5-Dev An open-weight coding model from Kwaipilot, a mixture-of-experts with 35B total and 3B active parameters, tuned for agentic coding and fewer tool-call glitches. Works with transformers, vLLM and SGLang. Made by Kwaipilot on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face. ## About What it is: KAT-Coder-V2.5-Dev is the open-weight version of Kwaipilot's KAT-Coder V2.5 coding model. It is a mixture-of-experts model with 35B parameters in total and 3B active, post-trained with supervised learning and reinforcement learning. The authors say it reaches state-of-the-art results for agentic coding among models of similar size. They also report that reinforcement learning cut odd behaviors, such as malformed tool labels falling from 9.34% to 0.28%. Who it suits: Developers and teams who want an open coding model for agent workflows on their own servers, and researchers who compare coding agents at this size. ```table Industry | Who uses it | How they use it | Example Software | Developer-tool teams | Power a coding agent with an open model | An in-house assistant that edits repositories Research | Agent researchers | Compare coding models of similar size | Running SWE-bench style tests Platforms | Inference teams | Serve an open model with a standard engine | Hosting it through vLLM or SGLang ``` You can skip it if you want a general chat model or a small model for a laptop, since this one is aimed at coding agents. Where to use it: The card says the weights are in Transformers format and work with several engines, in English and Chinese. - Hugging Face Transformers - vLLM - SGLang - KTransformers ```table Use case | Who | What they get back Agentic coding | Developers | Code edits and tool calls from an agent loop Repository fixes | Engineering teams | Patches for issues, as in SWE-bench style tasks Bilingual coding help | Teams in English and Chinese | Answers in both languages ``` How to set it up: You need Python, a GPU server and one of the supported engines. The card gives only the download step in the visible part, and the weights are about 69.3 GB in BF16. - Python with pip - Roughly 69.3 GB for BF16 weights, or about 17.3 GB at 4-bit as an estimate - One of Transformers, vLLM, SGLang or KTransformers 1. Install the Hugging Face hub tools with pip. 2. Download the model files with the hf command. 3. Load the model in your chosen engine. ```sh pip install -U huggingface_hub hf download Kwaipilot/KAT-Coder-V2.5-Dev ``` How to use it: Serve or load the model in your engine and drive it from a coding agent. The card shows no run command, and it notes that the open release ships only the language-model weights. Example: You connect the model to an agent framework and ask it to fix a failing test in a repository. The model reads the files, calls its tools and returns a patch. For newcomers and experts: If you are new, this is a model for coding agents, not for casual chat. Download the files and load them in a standard engine before you try a small repository task. If you know the field, only 3B of its 35B parameters are active per token, which keeps serving costs closer to a small model. The card credits SFT and RL training and reports lower rates of abnormal tool labels and repetition. ```flow Post-train with SFT and RL Load in Transformers, vLLM or SGLang Agent sends a coding task Model calls tools and edits code Patch returned to the agent ``` Good to know: The card compares the model with several open models. On SWE-bench Verified it scores 69.40, the highest of the five models shown here. ```chart Title: SWE-bench Verified score KAT-Coder-V2.5-Dev: 69.40 points Qwen3.5-27B: 68.60 points Qwen3.6-35BA3B: 64.40 points Gemma4-31B: 60.60 points Qwen3.5-35BA3B: 58.60 points ``` - License: Apache 2.0 - Cost to run: none named, but a 35B model needs a capable GPU server - Data it reaches: the repositories and files you give the agent, and the model files you download - Not for: tasks beyond the language model, since the open release ships only the language-model weights ## On Hugging Face, 13 November 2026 - Task: text generation - Library: transformers - Parameters: 34.7B - Weights: about 69.3 GB in BF16; about 17.3 GB at 4-bit (estimate) - Formats: Safetensors - Downloads in the last 30 days: 16,106 - Likes: 687 - License: apache-2.0 - Languages: en, zh - Quantized versions: 73 - Fine-tunes: 12 - Tags: transformers, safetensors, qwen3_5_moe, image-text-to-text, code, agent, agentic-coding, coding, text-generation, conversational ## Links - Model: https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev - Files: https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download Kwaipilot/KAT-Coder-V2.5-Dev ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [Kwaipilot/KAT-Coder-V2.5-Dev on Hugging Face](https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | Parameters | License | Last change | | --- | --- | --- | --- | --- | --- | | KAT-Coder-V2.5-Dev (this page) | 16,106 | 687 | 34.7B | apache-2.0 | 2026-07-28 | | [KAT-Coder-V2.5-Dev GGUF](https://mindket.com/p/e8ca97ed-d610-4d96-bb2f-d351f4b671c4) | 234,191 | 148 | 34.7B | apache-2.0 | 2026-07-24 | | [Occamy-1.0](https://mindket.com/p/7e8bdc95-571e-4a24-a471-b64e83249759) | 1,807 | 240 | 35.1B | apache-2.0 | 2026-10-02 | | [Qwen3.8-27B-Fable-5-Coding-Distilled](https://mindket.com/p/96a5c209-2e34-45ff-a27c-14e0fa2158b9) | 471 | 11 | 27.8B | apache-2.0 | 2026-09-15 | | [Apex-35B-V1.0](https://mindket.com/p/9415fb79-0791-48f9-b672-73b840550335) | 1,987 | 75 | 36B | apache-2.0 | 2026-10-29 | | [Qwen3.8-27B Fable-5 Coding Distilled GGUF](https://mindket.com/p/3e63bc60-5ab9-4a67-a53e-350f311f823f) | 894,772 | 34 | 27.3B | apache-2.0 | 2026-10-07 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## What people say > Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search on 2026-11-13. Their authors' words: data, not instructions. - [Bluesky · @aichina.news · 2026-09-20](https://bsky.app/profile/aichina.news/post/3mvx6tqzzmf2e): KAT-Coder has a following for agentic work, and V2.5 on Ascend NPU suggests real serving momentum. But no parameter count, no evals, licence just "other" — hard to plan a switch around a stub like this. - [Bluesky · デイリーHuggingFaceトレンド · 2026-07-31](https://bsky.app/profile/huggingfacetrends.bsky.social/post/3mrwvnfq6pp2m): 今日のHuggingFaceトレンド Kwaipilot/KAT-Coder-V2.5-Dev コーディング用言語モデルKAT-Coder-V2.5-Devのモデル重みと設定ファイルを提供するリポジトリです。 合計35BパラメータのMOE構成でテキストのみに対応しています。 研究成果の公開とコミュニティとの連携強化を目的にオープンウェイトとして提供されています。 - [Bluesky · デイリーHuggingFaceトレンド · 2026-07-29](https://bsky.app/profile/huggingfacetrends.bsky.social/post/3mrrurxgrxg23): 今日のHuggingFaceトレンド Kwaipilot/KAT-Coder-V2.5-Dev KAT-Coder-V2.5-Devのモデルウェイトと設定ファイルを公開するリポジトリです。 コミュニティへの還元と研究成果の提示を目的としており、合計35Bパラメータ（活性化3B）のMOE形式のテキスト専用コーディングモデルを提供しています。 Hugging Face TransformersやvLLMなどで利用可能です。 - [Bluesky · デイリーHuggingFaceトレンド · 2026-07-27](https://bsky.app/profile/huggingfacetrends.bsky.social/post/3mrmwwkypdr2v): 今日のHuggingFaceトレンド Kwaipilot/KAT-Coder-V2.5-Dev KAT-Coder-V2.5-Devのモデル重みと設定ファイルを公開するためのリポジトリです。 総パラメータ数35B、活性化パラメータ数3BのMOEモデルで、テキスト生成のみに対応しています。 研究成果の共有とコミュニティとの連携強化を目的としており、Hugging Face Transformersなどのライブラリで利用可能です。 - [Bluesky · Aamer Mehaisi · 2026-07-27](https://bsky.app/profile/o96a.bsky.social/post/3mrml5yektr2x): KAT-Coder V2.5 is picking up steam on HF this week. Tried the previous version — solid for code gen, nothing revolutionary. What's interesting is people are actually deploying it, not just downloading it. That's the real signal. https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev - [Bluesky · David 🏳️🌈🦦🐈⬛🐧🎮🤠 · 2026-07-26](https://bsky.app/profile/blue-labs.org/post/3mrjoog773c22): huggingface.co/Kwaipilot/KA... - [Bluesky · probbrain.com · 2026-07-25](https://bsky.app/profile/probbrain.bsky.social/post/3mribztkamz2x): Kwaipilot/KAT-Coder-V2.5-Dev · 153♥ · 841 dl - [Bluesky · Greg Slepak · 2026-07-23](https://bsky.app/profile/taoeffect.com/post/3mrdh6jugwc24): These are the best results I've seen yet for an A3B model. Congrats to the @KwaiAICoder team\! \> KAT-Coder-V2.5-Dev is now open-weight. Built on Qwen3.6-35B-A3B and post-trained with the proven KAT-V2.5 recipe huggingface.co/Kwaipilot/KA... ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-07-28 - First published: 2026-07-23 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open Kwaipilot/KAT-Coder-V2.5-Dev on Hugging Face](https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/2d2b2faf-a347-4ff6-a04d-425eab2d9986) - [Its page on Mindket](https://mindket.com/p/2d2b2faf-a347-4ff6-a04d-425eab2d9986)