diff --git a/src/maxtext/models/qwen3.py b/src/maxtext/models/qwen3.py
index 22f3335ba9..290847a44c 100644
--- a/src/maxtext/models/qwen3.py
+++ b/src/maxtext/models/qwen3.py
@@ -1305,6 +1305,7 @@ def __init__(self, config: Config, mesh: Mesh, quant: None | Quant = None, *, rn
           out_features_shape=1,
           use_bias=False,  # Qwen3-Next shared_expert_gate does not have a bias
           dtype=cfg.dtype,
+          weight_dtype=cfg.weight_dtype,
           kernel_init=max_initializers.nd_dense_init(cfg.dense_init_scale, "fan_in", "truncated_normal"),
           kernel_axes=("embed", None),
           matmul_precision=cfg.matmul_precision,