A vocabulary-trimmed version of Qwen/Qwen3-1.7B-Base: the byte-level BPE vocabulary is reduced from ~151.7k to 32768 tokens covering Korean + English, and the (tied) embedding matrix is sliced to match. No retraining — weights are copied verbatim for kept tokens, so on kept tokens the model is numerically identical to the original.

Method: Introduction to Trimming.

What changed

	Original	Trimmed
Parameters	1,720,574,976	1,476,518,912 (-14.2%)
Vocab size	151,669	32,768
Merges	151,387	32,486
Embedding	tied	tied (sliced)

All 256 byte-level tokens + all 26 special tokens are kept, plus the most frequent Korean/English tokens and their full BPE merge-derivation closure (so multi-byte Korean stays reachable).
Verified: lossless round-trip on KO/EN/code; teacher-forced logit equivalence max|Δ| = 0.0; greedy generation on natural Korean/English is token-for-token identical to the base model.

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "ceyda/Qwen3-1.7B-Base-trim-koen-32768"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo)

ids = tok("대한민국의 수도는", return_tensors="pt")
print(tok.decode(model.generate(**ids, max_new_tokens=20)[0], skip_special_tokens=True))

Limitations

Trimmed for Korean + English prose. Code and rare jargon still encode losslessly but may split into a few extra tokens. Text in other languages falls back to byte tokens (longer sequences).
This is a base (non-instruction-tuned) model.

Attribution

Derived from Qwen/Qwen3-1.7B-Base (Apache-2.0). Trimming method by Loïck Bourdois.