diff --git a/invokeai/app/invocations/latent.py b/invokeai/app/invocations/latent.py
index 743a648785..ed20704e41 100644
--- a/invokeai/app/invocations/latent.py
+++ b/invokeai/app/invocations/latent.py
@@ -22,7 +22,7 @@ from ...backend.stable_diffusion.diffusers_pipeline import (
 from ...backend.stable_diffusion.diffusion.shared_invokeai_diffusion import \
     PostprocessingSettings
 from ...backend.stable_diffusion.schedulers import SCHEDULER_MAP
-from ...backend.util.devices import choose_torch_device, torch_dtype
+from ...backend.util.devices import choose_torch_device, torch_dtype, choose_precision
 from ..models.image import ImageCategory, ImageField, ResourceOrigin
 from .baseinvocation import (BaseInvocation, BaseInvocationOutput,
                              InvocationConfig, InvocationContext)
@@ -492,7 +492,7 @@ class LatentsToImageInvocation(BaseInvocation):
     tiled: bool = Field(
         default=False,
         description="Decode latents by overlaping tiles(less memory consumption)")
-    fp32: bool = Field(False, description="Decode in full precision")
+    fp32: bool = Field(default=choose_precision(choose_torch_device())=='float32', description="Decode in full precision")
     metadata: Optional[CoreMetadata] = Field(default=None, description="Optional core metadata to be written to the image")
 
     # Schema customisation