HIT-SCIR/Chinese-Mixtral-8x7B
A Chinese-extended vocabulary large language model based on Mixtral-8x7B with vocabulary expansion and incremental pretraining.

Velocity · 7d
+0.7
★ / day
Trend
→steady
star history
This project extends the Mixtral-8x7B model with a Chinese-optimized vocabulary to improve encoding and decoding efficiency for Chinese text. It provides incremental pretraining code on large-scale open-source corpora and releases both LoRA adapter weights and fully merged model weights for download. The extended vocabulary significantly boosts the model’s Chinese language generation and comprehension capabilities.