mirror of
https://github.com/zenlm/zen-family.git
synced 2026-10-30 14:58:06 +00:00
zeekayandHanzo Dev
f6dce58b36
typescript 7.0.2 is the native compiler and exports no JavaScript API, so next 14.2.33 and eslint cannot read it. It comes in aliased as tsgo and does the typechecking; typescript stays on the 6 line for those. The site had no tsconfig, so this adds the one next writes for itself, with "types": ["node"] spelled out because TS7 no longer sweeps in every installed @types. 1 diagnostic before and after (fumadocs.config.ts sets a key GlobalConfig does not have). Co-authored-by: Hanzo Dev <[email protected]>
The Zen AI Model Family
Zen fine-tunes the best open-weight model of each era. Across this family that is the Qwen3 line from Alibaba Cloud — base, VL, Omni, Embedding, Reranker, TTS, ASR, Guard, Coder — with media models on permissive bases (Wan, FLUX, TRELLIS, YuE). Hanzo adds only identity training, agentic-data fine-tuning, and abliteration. There is no from-scratch Zen model.
Base models & attribution
- Language / VL / Omni / Coder / Guard / Embedding / Reranker / TTS / ASR: fine-tuned from Qwen3 (Alibaba Cloud, Apache-2.0). The reference language checkpoint fine-tunes Qwen/Qwen3-8B.
- Media (image / video / 3D / audio): permissive bases — FLUX.1-schnell (Apache-2.0), Wan2.2, TRELLIS (MIT), YuE.
Each model card on huggingface.co/zenlm lists its exact base_model and license.
Documentation
Quick start
pip install transformers torch accelerate
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("zenlm/zen-eco-4b-instruct")
tokenizer = AutoTokenizer.from_pretrained("zenlm/zen-eco-4b-instruct")
Citation
@article{zen2025,
title = {The Zen AI Model Family},
author = {Hanzo AI and Zoo Labs},
year = {2025}
}
License
Apache-2.0 — see LICENSE and NOTICE. Upstream Qwen3 is Apache-2.0.
Built by Hanzo AI and Zoo Labs Foundation