diff --git a/hls4ml/converters/__init__.py b/hls4ml/converters/__init__.py
index e38e82ac1..203e31a66 100644
--- a/hls4ml/converters/__init__.py
+++ b/hls4ml/converters/__init__.py
@@ -121,8 +121,16 @@ def convert_from_config(config):
         model = onnx_to_hls(yamlConfig)
     elif 'PytorchModel' in yamlConfig:
         model = pytorch_to_hls(yamlConfig)
-    else:
-        model = keras_v2_to_hls(yamlConfig)
+    elif 'KerasModel' in yamlConfig:
+        import keras
+
+        if keras.__version__ >= '3.0':
+            # Get fallback flags from config or use defaults
+            allow_da_fallback = yamlConfig.get('HLSConfig', {}).get('Model', {}).get('AllowDAFallback', True)
+            allow_v2_fallback = yamlConfig.get('HLSConfig', {}).get('Model', {}).get('AllowV2Fallback', True)
+            model = keras_v3_to_hls(yamlConfig, allow_da_fallback, allow_v2_fallback)
+        else:
+            model = keras_v2_to_hls(yamlConfig)
 
     return model
 
diff --git a/hls4ml/model/profiling.py b/hls4ml/model/profiling.py
index e122e3f46..81052a589 100644
--- a/hls4ml/model/profiling.py
+++ b/hls4ml/model/profiling.py
@@ -27,7 +27,7 @@
     __torch_profiling_enabled__ = False
 
 try:
-    import qkeras
+    from qkeras import QActivation
 
     __qkeras_profiling_enabled__ = True
 except ImportError:
@@ -37,7 +37,7 @@
 if __keras_profiling_enabled__:
     __keras_activations.append(keras.layers.Activation)
 if __qkeras_profiling_enabled__:
-    __keras_activations.append(qkeras.QActivation)
+    __keras_activations.append(QActivation)
 
 
 def get_unoptimized_hlsmodel(model):
diff --git a/test/pytest/generate_ci_yaml.py b/test/pytest/generate_ci_yaml.py
index fccf83de6..684abc051 100644
--- a/test/pytest/generate_ci_yaml.py
+++ b/test/pytest/generate_ci_yaml.py
@@ -28,7 +28,14 @@
 
 # Long-running tests will not be bundled with other tests
 LONGLIST = {'test_hgq_layers', 'test_hgq_players', 'test_qkeras', 'test_pytorch_api'}
-KERAS3_LIST = {'test_keras_v3_api', 'test_hgq2_mha', 'test_einsum_dense', 'test_qeinsum', 'test_multiout_onnx'}
+KERAS3_LIST = {
+    'test_keras_v3_api',
+    'test_hgq2_mha',
+    'test_einsum_dense',
+    'test_qeinsum',
+    'test_multiout_onnx',
+    'test_keras_v3_profiling',
+}
 
 # Test files to split by individual test cases
 # Value = chunk size per CI job
diff --git a/test/pytest/test_keras_v3_profiling.py b/test/pytest/test_keras_v3_profiling.py
new file mode 100644
index 000000000..f98e69963
--- /dev/null
+++ b/test/pytest/test_keras_v3_profiling.py
@@ -0,0 +1,141 @@
+"""Test numerical profiling with Keras v3 models."""
+
+import numpy as np
+import pytest
+
+try:
+    import keras
+
+    __keras_profiling_enabled__ = keras.__version__ >= '3.0'
+except ImportError:
+    __keras_profiling_enabled__ = False
+
+if __keras_profiling_enabled__:
+    from hls4ml.model.profiling import numerical
+
+
+def count_bars_in_figure(fig):
+    """Count the number of bars in all axes of a figure."""
+    count = 0
+    for ax in fig.get_axes():
+        count += len(ax.patches)
+    return count
+
+
+@pytest.mark.skipif(not __keras_profiling_enabled__, reason='Keras 3.0 or higher is required')
+def test_keras_v3_numerical_profiling_simple_model():
+    """Test numerical profiling with a simple Keras v3 Dense model."""
+    model = keras.Sequential(
+        [
+            keras.layers.Dense(20, input_shape=(10,), activation='relu'),
+            keras.layers.Dense(5, activation='softmax'),
+        ]
+    )
+    model.compile(optimizer='adam', loss='categorical_crossentropy')
+    # Build the model so weights are initialized
+    model.build((None, 10))
+
+    # Test profiling weights only
+    wp, _, _, _ = numerical(model)
+    assert wp is not None
+    # Should have 2 bars (one per layer, each showing weights and biases combined)
+    assert count_bars_in_figure(wp) == 2
+
+
+@pytest.mark.skipif(not __keras_profiling_enabled__, reason='Keras 3.0 or higher is required')
+def test_keras_v3_numerical_profiling_with_activations():
+    """Test numerical profiling with Keras v3 model including activations."""
+    # Use functional API instead of Sequential to ensure input layer is properly defined
+    inputs = keras.Input(shape=(10,))
+    x = keras.layers.Dense(20, activation='relu')(inputs)
+    outputs = keras.layers.Dense(5)(x)
+    model = keras.Model(inputs=inputs, outputs=outputs)
+    model.compile(optimizer='adam', loss='mse')
+
+    # Generate test data
+    X_test = np.random.rand(100, 10).astype(np.float32)
+
+    # Test profiling with activations
+    wp, _, ap, _ = numerical(model, X=X_test)
+    assert wp is not None
+    assert ap is not None
+
+
+@pytest.mark.skipif(not __keras_profiling_enabled__, reason='Keras 3.0 or higher is required')
+def test_keras_v3_numerical_profiling_conv_model():
+    """Test numerical profiling with a Keras v3 Conv model."""
+    model = keras.Sequential(
+        [
+            keras.layers.Conv2D(16, (3, 3), activation='relu', input_shape=(28, 28, 1)),
+            keras.layers.MaxPooling2D((2, 2)),
+            keras.layers.Flatten(),
+            keras.layers.Dense(10, activation='softmax'),
+        ]
+    )
+    model.compile(optimizer='adam', loss='categorical_crossentropy')
+    # Build the model so weights are initialized
+    model.build((None, 28, 28, 1))
+
+    # Test profiling weights
+    wp, _, _, _ = numerical(model)
+    assert wp is not None
+    # Conv layer has 1 bar, Dense layer has 1 bar = 2 bars total
+    assert count_bars_in_figure(wp) == 2
+
+
+@pytest.mark.skipif(not __keras_profiling_enabled__, reason='Keras 3.0 or higher is required')
+@pytest.mark.skip(reason='convert_from_config needs update for Keras v3 model serialization format')
+def test_keras_v3_numerical_profiling_with_hls_model():
+    """Test numerical profiling with both Keras v3 model and hls4ml model."""
+    import hls4ml
+
+    # Use functional API to ensure input layer is properly defined
+    inputs = keras.Input(shape=(8,))
+    x = keras.layers.Dense(16, activation='relu')(inputs)
+    outputs = keras.layers.Dense(4, activation='softmax')(x)
+    model = keras.Model(inputs=inputs, outputs=outputs)
+    model.compile(optimizer='adam', loss='categorical_crossentropy')
+
+    # Generate test data
+    X_test = np.random.rand(100, 8).astype(np.float32)
+
+    # Create hls4ml model
+    config = hls4ml.utils.config_from_keras_model(model, granularity='name')
+    hls_model = hls4ml.converters.convert_from_keras_model(
+        model,
+        hls_config=config,
+        output_dir='/tmp/test_keras_v3_profiling_hls',
+        backend='Vivado',
+        allow_da_fallback=True,
+        allow_v2_fallback=True,
+    )
+
+    # Test profiling with both models
+    wp, wph, ap, aph = numerical(model, hls_model=hls_model, X=X_test)
+
+    assert wp is not None  # Keras model weights (before optimization)
+    assert wph is not None  # HLS model weights (after optimization)
+    assert ap is not None  # Keras model activations (before optimization)
+    assert aph is not None  # HLS model activations (after optimization)
+
+
+@pytest.mark.skipif(not __keras_profiling_enabled__, reason='Keras 3.0 or higher is required')
+def test_keras_v3_numerical_profiling_batch_norm():
+    """Test numerical profiling with Keras v3 model containing BatchNormalization."""
+    model = keras.Sequential(
+        [
+            keras.layers.Dense(20, input_shape=(10,)),
+            keras.layers.BatchNormalization(),
+            keras.layers.Activation('relu'),
+            keras.layers.Dense(5, activation='softmax'),
+        ]
+    )
+    model.compile(optimizer='adam', loss='categorical_crossentropy')
+    # Build the model so weights are initialized
+    model.build((None, 10))
+
+    # Test profiling weights
+    wp, _, _, _ = numerical(model)
+    assert wp is not None
+    # Dense has 1 bar, BatchNorm has 1 bar, second Dense has 1 bar = 3 bars
+    assert count_bars_in_figure(wp) == 3