Qwen3.5-4B-oQ3.5

This model was quantized using oQ mixed-precision quantization.

Quantization details

• Model type: qwen3_5
• Bits: 3
• Group size: 64
• Format: MLX safetensors

Benchmark

Model	Benchmark	Accuracy	Correct	Total	Time(s)
Qwen3.5-4B-oQ3.5	MMLU	47.3%	142	300	399.3
Qwen3.5-4B-oQ3.5+	MMLU	50.0%	150	300	402.5
Qwen3.5-4B-oQ4	MMLU	67.3%	202	300	515.5
Qwen3.5-4B-oQ3.5	JMMLU	57.3%	172	300	96.2
Qwen3.5-4B-oQ3.5+	JMMLU	56.3%	169	300	94.2
Qwen3.5-4B-oQ4	JMMLU	62.7%	188	300	121.5