Qwen3-Embedding-4B-W4A16-G128

GPTQ Quantized Qwen/Qwen3-Embedding-4B with THUIR/T2Ranking and m-a-p/COIG-CQIA for calibration set.

What's the benefit?

VRAM Usage: 17430M -> 11000M (w/o FA2).

What's the cost?

~0.72% lost in C-MTEB.

Evaluation performed with official code.

C-MTEB	Param.	Mean(Task)	Mean(Type)	Class.	Clust.	Pair Class.	Rerank.	Retr.	STS
multilingual-e5-large-instruct	0.6B	58.08	58.24	69.80	48.23	64.52	57.45	63.65	45.81
bge-multilingual-gemma2	9B	67.64	68.52	75.31	59.30	86.67	68.28	73.73	55.19
gte-Qwen2-1.5B-instruct	1.5B	67.12	67.79	72.53	54.61	79.5	68.21	71.86	60.05
gte-Qwen2-7B-instruct	7.6B	71.62	72.19	75.77	66.06	81.16	69.24	75.70	65.20
ritrieve_zh_v1	0.3B	72.71	73.85	76.88	66.5	85.98	72.86	76.97	63.92
Qwen3-Embedding-4B	4B	72.27	73.51	75.46	77.89	83.34	66.05	77.03	61.26
This Model	4B-W4A16	71.75	73.05	75.43	77.51	83.04	65.73	76.15	60.47

How to use it?

pip install compressed-tensors optimum and auto-gptq / gptqmodel, then goto the official usage guide.