# Zyphra converts pretrained models with ZHA using 81,920 trainable parameters

> Original title: There is no need to train from scratch.

- Source: Zyphra
- Published: 2026-11-02T22:04:53.000Z
- Bad Influence: https://badinfluence.news/items/zaahw8f1y6mgdiv2uvrm1l7hy
- Original: https://x.com/ZyphraAI/status/2117377434211496426

## Summary

Zyphra says it converts existing pretrained models through distillation using only a few billion tokens while keeping the base model frozen, so no training from scratch is needed. For Qwen3.6-35B-A3B, its ZHA method adds only 81,920 trainable parameters, about 0.00024% of the model.

## Post

There is no need to train from scratch.

We convert existing pretrained models with just a few billion tokens of distillation while freezing the base model. For Qwen3.6-35B-A3B, ZHA adds only 81,920 trainable parameters (~0.00024% of the model).