What are the hardware requirements for KAT Coder V2.5 Dev?

KAT Coder V2.5 Dev requires approximately 69.7 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed KAT Coder V2.5 Dev and how large is it?

KAT Coder V2.5 Dev is developed by Kwaipilot, a model with 35b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate KAT Coder V2.5 Dev?

KAT Coder V2.5 Dev is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

KAT Coder V2.5 Dev by Kwaipilot — VRAM 69.7GB

Name: KAT Coder V2.5 Dev
Author: Kwaipilot

KAT Coder V2.5 Dev is an open-source language model by Kwaipilot. Features: 35b LLM, VRAM: 69.7GB, License: apache-2.0, Code Generating, LLM Explorer Score: 0.42.

Arxiv:2607.05471 Agent Agentic-coding Code Codegen Coding Conversational En Endpoints compatible Image-text-to-text Moe Qwen3 5 moe Region:us Safetensors Sharded Tensorflow Zh

Model Card on HF 🤗: https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev

KAT Coder V2.5 Dev Parameters and Internals

LLM Name	KAT Coder V2.5 Dev
Repository 🤗	https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev
Base Model(s)	Qwen3.6-35B-A3B /Qwen3.6-35B-A3B
Model Size	35b
Required VRAM	69.7 GB
Updated	2026-08-08
Maintainer	Kwaipilot
Model Type	qwen3_5_moe
Model Files	5.4 GB: 0-of-13 5.4 GB: 1-of-13 5.4 GB: 2-of-13 5.4 GB: 3-of-13 5.4 GB: 4-of-13 5.4 GB: 5-of-13 5.4 GB: 6-of-13 5.4 GB: 7-of-13 5.4 GB: 8-of-13 5.4 GB: 9-of-13 5.4 GB: 10-of-13 5.4 GB: 11-of-13 4.9 GB: 12-of-13
Supported Languages	en zh
Generates Code	Yes
Model Architecture	Qwen3_5MoeForConditionalGeneration
License	apache-2.0
Model Max Length	262144
Transformers Version	4.57.1
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Errors	replace

Rank the KAT Coder V2.5 Dev capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Quantized Models of the KAT Coder V2.5 Dev

Model	Likes	Downloads	VRAM
...T Coder V2.5 Dev MTP APEX GGUF	8	10166	1 GB
... UNCENSORED PHILADELPHIA CLASS	1	2804	21 GB
KAT Coder V2.5 Dev OptiQ 4bit	3	114	21 GB

Best Alternatives to KAT Coder V2.5 Dev

Best Alternatives	Context / RAM	Downloads	Likes
...iel Coder 35B A3B MLX OQ6e MTP	0K / 29.5 GB	569	3
Tiel Coder 35B A3B MLX OQ4e	0K / 21 GB	565	7
...iel Coder 35B A3B MLX OQ6e MTP	0K / 29.5 GB	283	4
Qwopus KAT Coder 35B Merged	0K / 137 GB	274	1
...iel Coder 35B A3B MLX OQ4e MTP	0K / 21 GB	115	3
KAT Coder V2.5 Dev VL OQ8e Mtp	0K / 38.4 GB	298	2
Tiel Coder 35B A3B MLX OQ6e	0K / 29.5 GB	84	1
KAT Coder V2.5 Dev OQ6e	0K / 28.5 GB	142	1
KAT Coder V2.5 Dev VL OQ4e Mtp	0K / 21.5 GB	149	1
Qwopus3.6 35B A3B Coder Bf16	0K / 69.9 GB	227	1