Back to Kwaipilot
Kwaipilot /

KA
KAT-Coder-V2.5-Dev

Open WeightsChat & ReasoningtextUpdated July 23, 2026

KAT-Coder-V2.5-Dev

Model Overview

KAT-Coder-V2.5-Dev is a 34.7B-parameter Mixture-of-Experts text generation model developed by Kwaipilot. Built on the Qwen3_5MoeForConditionalGeneration architecture. Supports en, zh. Released on 2026-07-23.


📊 Quick Specs

Specification Table
SpecificationValue
Parameters34.7B
ArchitectureQwen3_5MoeForConditionalGeneration
Tasktext generation
Modalitytext
LicenseAPACHE-2.0
Frameworktransformers
MoEYes
Languagesen, zh

✨ Key Features

  • 34.7B parameters with sparse MoE architecture for efficient inference
  • Built on Qwen3_5MoeForConditionalGeneration architecture (transformers)
  • Primary task: text generation (text modality)
  • Language support: en, zh
  • Open-weights under APACHE-2.0 license — self-hostable and fine-tunable

📈 Community Adoption

  • 126 likes on Hugging Face
  • 396 downloads on Hugging Face

🔗 Resources


📜 License & Access

APACHE-2.0 — Open-weights model available for download, fine-tuning, and self-hosted deployment.

Key Features

34.7B parameters with sparse MoE architecture for efficient inference

Feature 01

Built on Qwen3_5MoeForConditionalGeneration architecture (transformers)

Feature 02

Primary task: text generation (text modality)

Feature 03

Language support: en, zh

Feature 04

Open-weights under APACHE-2.0 license — self-hostable and fine-tunable

Feature 05

You might also want to compare

Verified Sources

Tags

open-weightstrendinghuggingfacemoetransformerstext-generation

Model Specs

open-weights

Parameters

34.7B

Context Window

unknown

License

APACHE-2.0

Deployment

self-hostable

Resources & Links

Curator Notes

Auto-ingested from HuggingFace Trending on 2026-07-25. Architecture: Qwen3_5MoeForConditionalGeneration. Params: 34.7B. License: APACHE-2.0.

Compare Specs

Compare parameters, context windows, modalities, and benchmark scores of this model side-by-side with others.

Compare Model