diff --git a/.github/workflows/conda-package.yml b/.github/workflows/conda-package.yml
index 5478023849..ea5ef1f89f 100644
--- a/.github/workflows/conda-package.yml
+++ b/.github/workflows/conda-package.yml
@@ -25,6 +25,10 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
     steps:
       - name: Cancel Previous Runs
         uses: styfle/cancel-workflow-action@3155a141048f8f89c06b4cdae32e7853e97536bc # 0.13.0
@@ -45,9 +49,9 @@ jobs:
         with:
           path: ~/.conda/pkgs
           key:
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-${{hashFiles('**/meta.yaml') }}
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-${{hashFiles('**/meta.yaml') }}
           restore-keys: |
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-
             ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-
       - name: Add conda to system path
         run: echo $CONDA/bin >> $GITHUB_PATH
@@ -61,22 +65,31 @@ jobs:
         run: |
           # use bootstrap channel to pull NumPy linked with OpenBLAS
           CHANNELS="-c conda-forge --override-channels"
-          VERSIONS="--python ${{ matrix.python }} --numpy 2.0"
           TEST="--no-test"
-          conda build \
-            $TEST \
-            $VERSIONS \
-            $CHANNELS \
-            conda-recipe
+          if [ -n "${{ matrix.python_spec }}" ]; then
+            conda build \
+              $TEST \
+              --python "${{ matrix.python_spec }}" \
+              --numpy 2.0 \
+              $CHANNELS \
+              conda-recipe
+          else
+            conda build \
+              $TEST \
+              --python ${{ matrix.python }} \
+              --numpy 2.0 \
+              $CHANNELS \
+              conda-recipe
+          fi
       - name: Upload artifact
         uses: actions/upload-artifact@v6.0.0
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
           path: /usr/share/miniconda/conda-bld/linux-64/${{ env.PACKAGE_NAME }}-*.conda
       - name: Upload wheels artifact
         uses: actions/upload-artifact@v6.0.0
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
           path: ${{ env.WHEELS_OUTPUT_FOLDER }}${{ env.PACKAGE_NAME }}-*.whl
 
   build_windows:
@@ -86,6 +99,10 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
     steps:
       - name: Cancel Previous Runs
         uses: styfle/cancel-workflow-action@3155a141048f8f89c06b4cdae32e7853e97536bc # 0.13.0
@@ -100,15 +117,15 @@ jobs:
         with:
           miniforge-variant: Miniforge3
           miniforge-version: latest
-          activate-environment: build
+          auto-activate: true
+          activate-environment: base
           channels: conda-forge
           conda-remove-defaults: true
           python-version: ${{ matrix.python }}
 
       - name: Install conda build
         run: |
-          conda activate
-          conda install -y conda-build
+          conda install -n base -y conda-build
           conda list -n base
 
       - name: Cache conda packages
@@ -118,9 +135,9 @@ jobs:
         with:
           path: /home/runner/conda_pkgs_dir
           key:
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-${{hashFiles('**/meta.yaml') }}
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-${{hashFiles('**/meta.yaml') }}
           restore-keys: |
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-
             ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-
 
       - name: Store conda paths as envs
@@ -133,20 +150,24 @@ jobs:
         env:
           OVERRIDE_INTEL_IPO: 1   # IPO requires more resources that GH actions VM provides
         run: |
-          conda activate
           # TODO: roll back use of Intel channel when 2025.1 is available on conda-forge
-          conda build --no-test --python ${{ matrix.python }} --numpy 2.0 -c ${{ env.INTEL_CHANNEL }} -c conda-forge --override-channels conda-recipe
+          $PYTHON_SPEC = "${{ matrix.python_spec }}"
+          if ($PYTHON_SPEC -ne "") {
+            conda build --no-test --python "$PYTHON_SPEC" --numpy 2.0 -c ${{ env.INTEL_CHANNEL }} -c conda-forge --override-channels conda-recipe
+          } else {
+            conda build --no-test --python ${{ matrix.python }} --numpy 2.0 -c ${{ env.INTEL_CHANNEL }} -c conda-forge --override-channels conda-recipe
+          }
 
       - name: Upload artifact
         uses: actions/upload-artifact@v6.0.0
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
           path: ${{ env.CONDA_BLD }}${{ env.PACKAGE_NAME }}-*.conda
 
       - name: Upload wheels artifact
         uses: actions/upload-artifact@v6.0.0
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
           path: ${{ env.WHEELS_OUTPUT_FOLDER }}${{ env.PACKAGE_NAME }}-*.whl
 
   test_linux:
@@ -157,6 +178,12 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
+            experimental: false
+            runner: ubuntu-22.04
         experimental: [false]
         runner: [ubuntu-22.04]
     continue-on-error: ${{ matrix.experimental }}
@@ -171,7 +198,7 @@ jobs:
       - name: Download artifact
         uses: actions/download-artifact@v7
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
       - name: Add conda to system path
         run: echo $CONDA/bin >> $GITHUB_PATH
       - name: Install conda-index
@@ -192,7 +219,11 @@ jobs:
         run: |
           CHANNELS="-c $GITHUB_WORKSPACE/channel ${{ env.CHANNELS }}"
           export PACKAGE_VERSION=$(python -c "${VER_SCRIPT1} ${VER_SCRIPT2}")
-          conda create -n ${{ env.TEST_ENV_NAME }} $PACKAGE_NAME=${PACKAGE_VERSION} python=${{ matrix.python }} $CHANNELS --only-deps --dry-run > lockfile
+          PYTHON_SPEC="${{ matrix.python_spec }}"
+          if [ -z "${PYTHON_SPEC}" ]; then
+            PYTHON_SPEC="${{ matrix.python }}"
+          fi
+          conda create -n ${{ env.TEST_ENV_NAME }} $PACKAGE_NAME=${PACKAGE_VERSION} python="${PYTHON_SPEC}" $CHANNELS --only-deps --dry-run > lockfile
           cat lockfile
       - name: Set pkgs_dirs
         run: |
@@ -204,16 +235,20 @@ jobs:
         with:
           path: ~/.conda/pkgs
           key:
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-${{hashFiles('lockfile') }}
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-${{hashFiles('lockfile') }}
           restore-keys: |
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-
             ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-
       - name: Install dpctl
         run: |
           export CHANNELS="-c $GITHUB_WORKSPACE/channel ${{ env.CHANNELS }}"
-          export TEST_DEPENDENCIES="pytest pytest-cov cython setuptools"
+          export TEST_DEPENDENCIES="pytest cython setuptools"
           export PACKAGE_VERSION=$(python -c "${VER_SCRIPT1} ${VER_SCRIPT2}")
-          conda create -n ${{ env.TEST_ENV_NAME }} $PACKAGE_NAME=${PACKAGE_VERSION} ${TEST_DEPENDENCIES} python=${{ matrix.python }} ${CHANNELS}
+          PYTHON_SPEC="${{ matrix.python_spec }}"
+          if [ -z "${PYTHON_SPEC}" ]; then
+            PYTHON_SPEC="${{ matrix.python }}"
+          fi
+          conda create -n ${{ env.TEST_ENV_NAME }} $PACKAGE_NAME=${PACKAGE_VERSION} ${TEST_DEPENDENCIES} python="${PYTHON_SPEC}" ${CHANNELS}
           # Test installed packages
           conda list -n ${{ env.TEST_ENV_NAME }}
       - name: Smoke test
@@ -253,6 +288,12 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
+            experimental: false
+            runner: windows-latest
         experimental: [false]
         runner: [windows-latest]
     continue-on-error: ${{ matrix.experimental }}
@@ -272,7 +313,7 @@ jobs:
       - name: Download artifact
         uses: actions/download-artifact@v7
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
 
       - uses: conda-incubator/setup-miniconda@v3
         with:
@@ -329,7 +370,9 @@ jobs:
           FOR /F "tokens=* USEBACKQ" %%F IN (`python -c "%SCRIPT%"`) DO (
              SET PACKAGE_VERSION=%%F
           )
-          conda install -n ${{ env.TEST_ENV_NAME }} ${{ env.PACKAGE_NAME }}=%PACKAGE_VERSION% python=${{ matrix.python }} -c ${{ env.workdir }}/channel ${{ env.CHANNELS }} --only-deps --dry-run > lockfile
+           SET "PYTHON_SPEC=${{ matrix.python_spec }}"
+           IF "%PYTHON_SPEC%"=="" SET "PYTHON_SPEC=${{ matrix.python }}"
+          conda install -n ${{ env.TEST_ENV_NAME }} ${{ env.PACKAGE_NAME }}=%PACKAGE_VERSION% python="%PYTHON_SPEC%" -c ${{ env.workdir }}/channel ${{ env.CHANNELS }} --only-deps --dry-run > lockfile
 
       - name: Display lockfile content
         shell: pwsh
@@ -342,9 +385,9 @@ jobs:
         with:
           path: /home/runner/conda_pkgs_dir
           key:
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-${{hashFiles('lockfile') }}
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-${{hashFiles('lockfile') }}
           restore-keys: |
-            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-
+            ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-python-${{ matrix.python }}-spec-${{ matrix.python == '3.14' && matrix.python_spec == '' && 't' || '' }}-
             ${{ runner.os }}-conda-${{ env.CACHE_NUMBER }}-
 
       - name: Install opencl_rt
@@ -362,8 +405,10 @@ jobs:
           FOR /F "tokens=* USEBACKQ" %%F IN (`python -c "%SCRIPT%"`) DO (
              SET PACKAGE_VERSION=%%F
           )
-          SET TEST_DEPENDENCIES=pytest"<8" pytest-cov cython setuptools
-          conda install -n ${{ env.TEST_ENV_NAME }} ${{ env.PACKAGE_NAME }}=%PACKAGE_VERSION% %TEST_DEPENDENCIES% python=${{ matrix.python }} -c ${{ env.workdir }}/channel ${{ env.CHANNELS }}
+           SET "PYTHON_SPEC=${{ matrix.python_spec }}"
+           IF "%PYTHON_SPEC%"=="" SET "PYTHON_SPEC=${{ matrix.python }}"
+          SET TEST_DEPENDENCIES=pytest"<8" cython setuptools
+           conda install -n ${{ env.TEST_ENV_NAME }} ${{ env.PACKAGE_NAME }}=%PACKAGE_VERSION% %TEST_DEPENDENCIES% python="%PYTHON_SPEC%" -c ${{ env.workdir }}/channel ${{ env.CHANNELS }}
 
       - name: Report content of test environment
         shell: cmd /C CALL {0}
@@ -424,16 +469,20 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
     steps:
       - name: Download conda artifact
         uses: actions/download-artifact@v7
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
 
       - name: Download wheel artifact
         uses: actions/download-artifact@v7
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
 
       - name: Install anaconda-client
         run: conda install anaconda-client -c conda-forge --override-channels
@@ -468,24 +517,28 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
     steps:
       - name: Download artifact
         uses: actions/download-artifact@v7
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
 
       - name: Download wheel artifact
         uses: actions/download-artifact@v7
         with:
-          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python }}
+          name: ${{ env.PACKAGE_NAME }} ${{ runner.os }} Wheels Python ${{ matrix.python == '3.14' && (matrix.python_spec != '' && '3.14' || '3.14t') || matrix.python }}
 
       - uses: conda-incubator/setup-miniconda@v3
         with:
           miniforge-version: latest
           channels: conda-forge
           conda-remove-defaults: true
-          auto-activate-base: true
-          activate-environment: ""
+          auto-activate: true
+          activate-environment: base
 
       - name: Install anaconda-client
         run: conda install anaconda-client -c conda-forge --override-channels
diff --git a/.github/workflows/run-tests-from-dppy-bits.yaml b/.github/workflows/run-tests-from-dppy-bits.yaml
index 0ee861aea1..7a1768310c 100644
--- a/.github/workflows/run-tests-from-dppy-bits.yaml
+++ b/.github/workflows/run-tests-from-dppy-bits.yaml
@@ -28,6 +28,10 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
         experimental: [false]
         runner: [ubuntu-22.04, ubuntu-24.04]
     continue-on-error: ${{ matrix.experimental }}
@@ -47,7 +51,11 @@ jobs:
 
       - name: Install dpctl
         run: |
-          conda create -n ${{ env.TEST_ENV_NAME }} -c dppy/label/dev ${{ env.CHANNELS }} dpctl pytest pytest-cov cython setuptools c-compiler cxx-compiler
+          PYTHON_SPEC="${{ matrix.python_spec }}"
+          if [ -z "${PYTHON_SPEC}" ]; then
+            PYTHON_SPEC="${{ matrix.python }}"
+          fi
+          conda create -n ${{ env.TEST_ENV_NAME }} -c dppy/label/dev ${{ env.CHANNELS }} dpctl pytest cython setuptools c-compiler cxx-compiler python="${PYTHON_SPEC}"
 
       - name: Smoke test
         run: |
@@ -79,6 +87,10 @@ jobs:
     strategy:
       matrix:
         python: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python_spec: ['']
+        include:
+          - python: '3.14'
+            python_spec: '3.14.* *_cp314'
         experimental: [false]
         runner: [windows-latest]
 
@@ -106,7 +118,9 @@ jobs:
 
       - name: Install dpctl
         run: |
-          conda install -n ${{ env.TEST_ENV_NAME }} -c dppy/label/dev ${{ env.CHANNELS }} dpctl pytest pytest-cov cython setuptools c-compiler cxx-compiler
+          SET "PYTHON_SPEC=${{ matrix.python_spec }}"
+          IF "%PYTHON_SPEC%"=="" SET "PYTHON_SPEC=${{ matrix.python }}"
+          conda install -n ${{ env.TEST_ENV_NAME }} -c dppy/label/dev ${{ env.CHANNELS }} dpctl pytest cython setuptools c-compiler cxx-compiler python="%PYTHON_SPEC%"
 
       # intel-opencl-rt is not being installed when running conda install dpctl, so do it manually
       - name: Install intel-opencl-rt
diff --git a/conda-recipe/meta.yaml b/conda-recipe/meta.yaml
index a5ad3bae93..99bfbfadbd 100644
--- a/conda-recipe/meta.yaml
+++ b/conda-recipe/meta.yaml
@@ -28,7 +28,6 @@ requirements:
         - {{ compiler('dpcpp') }} >={{ required_compiler_version }}
     host:
         - python
-        - python-gil     # [py>=314]
         - pip >=24.0
         - level-zero-devel >=1.16
         - pybind11 >=2.12
@@ -52,7 +51,6 @@ requirements:
         - tomli # [py<311]
     run:
         - python
-        - python-gil     # [py>=314]
         - {{ pin_compatible('intel-sycl-rt', min_pin='x.x', max_pin='x') }}
         - {{ pin_compatible('intel-cmplr-lib-rt', min_pin='x.x', max_pin='x') }}
         - numpy
@@ -65,7 +63,6 @@ test:
         - cython
         - setuptools
         - pytest
-        - pytest-cov
 
 about:
     home: https://github.com/IntelPython/dpctl.git
diff --git a/conda-recipe/run_test.sh b/conda-recipe/run_test.sh
index 504977258b..de137e0caa 100644
--- a/conda-recipe/run_test.sh
+++ b/conda-recipe/run_test.sh
@@ -4,11 +4,4 @@ set -e
 
 ${PYTHON} -c "import dpctl; print(dpctl.__version__)"
 ${PYTHON} -m dpctl -f
-# don't use coverage for Python 3.13 due to crashes related to
-# Cython >= 3.1.0 and Python >= 3.13
-# TODO: remove if crash is triaged
-if ${PYTHON} --version 2>&1 | grep -q '^Python 3\.13'; then
-    ${PYTHON} -m pytest -q -ra --disable-warnings --pyargs dpctl -vv
-else
-    ${PYTHON} -m pytest -q -ra --disable-warnings --cov dpctl --cov-report term-missing --pyargs dpctl -vv
-fi
+${PYTHON} -m pytest -q -ra --disable-warnings --pyargs dpctl -vv
diff --git a/dpctl/_diagnostics.pyx b/dpctl/_diagnostics.pyx
index 3723c23493..2d5879dfea 100644
--- a/dpctl/_diagnostics.pyx
+++ b/dpctl/_diagnostics.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ Implements developer utilities.
 """
@@ -60,6 +61,10 @@ def _shutdown_logger():
 def syclinterface_diagnostics(verbosity="warning", log_dir=None):
     """Context manager that activate verbosity of DPCTLSyclInterface
     function calls.
+
+    .. warning::
+        This context manager modifies the ``DPCTL_VERBOSITY`` environment
+        variable and should only be used from a single thread.
     """
     _allowed_verbosity = ["warning", "error"]
     if verbosity not in _allowed_verbosity:
diff --git a/dpctl/_sycl_context.pyx b/dpctl/_sycl_context.pyx
index 1cc1a5ac3b..7a0074db61 100644
--- a/dpctl/_sycl_context.pyx
+++ b/dpctl/_sycl_context.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ Implements SyclContext Cython extension type.
 """
diff --git a/dpctl/_sycl_device.pyx b/dpctl/_sycl_device.pyx
index 419ed2b9fb..25b616c9ee 100644
--- a/dpctl/_sycl_device.pyx
+++ b/dpctl/_sycl_device.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ Implements SyclDevice Cython extension type.
 """
diff --git a/dpctl/_sycl_device_factory.pyx b/dpctl/_sycl_device_factory.pyx
index 1689249bc9..4579b420ee 100644
--- a/dpctl/_sycl_device_factory.pyx
+++ b/dpctl/_sycl_device_factory.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ This module implements several device creation helper functions:
 
@@ -438,12 +439,27 @@ cdef class _DefaultDeviceCache:
         return _copy
 
 
+# no default, as would share a single mutable instance across threads and
+# concurrent access to the cache would not be thread-safe. Using ContextVar
+# without a default ensures each context gets its own instance.
 _global_default_device_cache = ContextVar(
     "global_default_device_cache",
-    default=_DefaultDeviceCache()
 )
 
 
+cdef _DefaultDeviceCache _get_default_device_cache():
+    """
+    Factory function to get or create a default device cache for the current
+    context
+    """
+    try:
+        return _global_default_device_cache.get()
+    except LookupError:
+        cache = _DefaultDeviceCache()
+        _global_default_device_cache.set(cache)
+        return cache
+
+
 cpdef SyclDevice _cached_default_device():
     """Returns a cached device selected by default selector.
 
@@ -452,7 +468,7 @@ cpdef SyclDevice _cached_default_device():
             A cached default-selected SYCL device.
 
     """
-    cdef _DefaultDeviceCache _cache = _global_default_device_cache.get()
+    cdef _DefaultDeviceCache _cache = _get_default_device_cache()
     d_, changed_ = _cache.get_or_create()
     if changed_:
         _global_default_device_cache.set(_cache)
diff --git a/dpctl/_sycl_event.pyx b/dpctl/_sycl_event.pyx
index 8766408644..6aa6a49d51 100644
--- a/dpctl/_sycl_event.pyx
+++ b/dpctl/_sycl_event.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ Implements SyclEvent Cython extension type.
 """
diff --git a/dpctl/_sycl_platform.pyx b/dpctl/_sycl_platform.pyx
index 7ceb725083..8a2f7a6bee 100644
--- a/dpctl/_sycl_platform.pyx
+++ b/dpctl/_sycl_platform.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ Implements SyclPlatform Cython extension type.
 """
diff --git a/dpctl/_sycl_queue.pxd b/dpctl/_sycl_queue.pxd
index 9469415cb8..c72cfb78e2 100644
--- a/dpctl/_sycl_queue.pxd
+++ b/dpctl/_sycl_queue.pxd
@@ -43,6 +43,7 @@ cdef public api class _SyclQueue [
     cdef DPCTLSyclQueueRef _queue_ref
     cdef SyclContext _context
     cdef SyclDevice _device
+    cdef object __weakref__
 
 
 cdef public api class SyclQueue (_SyclQueue) [
diff --git a/dpctl/_sycl_queue.pyx b/dpctl/_sycl_queue.pyx
index 48c882b860..c13f1f6d1e 100644
--- a/dpctl/_sycl_queue.pyx
+++ b/dpctl/_sycl_queue.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """ Implements SyclQueue Cython extension type.
 """
diff --git a/dpctl/_sycl_queue_manager.pyx b/dpctl/_sycl_queue_manager.pyx
index bf720e9402..274de15a29 100644
--- a/dpctl/_sycl_queue_manager.pyx
+++ b/dpctl/_sycl_queue_manager.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 import logging
 from contextvars import ContextVar
@@ -74,12 +75,27 @@ cdef class _DeviceDefaultQueueCache:
         return _copy
 
 
+# no default, as would share a single mutable instance across threads and
+# concurrent access to the cache would not be thread-safe. Using ContextVar
+# without a default ensures each context gets its own instance.
 _global_device_queue_cache = ContextVar(
     "global_device_queue_cache",
-    default=_DeviceDefaultQueueCache()
 )
 
 
+cdef _DeviceDefaultQueueCache _get_device_queue_cache():
+    """
+    Factory function to get or create a default device queue cache for the
+    current context
+    """
+    try:
+        return _global_device_queue_cache.get()
+    except LookupError:
+        cache = _DeviceDefaultQueueCache()
+        _global_device_queue_cache.set(cache)
+        return cache
+
+
 cpdef object get_device_cached_queue(object key):
     """Returns a cached queue associated with given device.
 
@@ -96,7 +112,7 @@ cpdef object get_device_cached_queue(object key):
         TypeError: If the input key is not one of the accepted types.
 
     """
-    _cache = _global_device_queue_cache.get()
+    _cache = _get_device_queue_cache()
     q_, changed_ = _cache.get_or_create(key)
     if changed_:
         _global_device_queue_cache.set(_cache)
diff --git a/dpctl/memory/_memory.pyx b/dpctl/memory/_memory.pyx
index 0cc6394741..f59025e55f 100644
--- a/dpctl/memory/_memory.pyx
+++ b/dpctl/memory/_memory.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """This file implements Python buffer protocol using Sycl USM shared and host
 allocators. The USM device allocator is also exposed through this module for
diff --git a/dpctl/program/_program.pyx b/dpctl/program/_program.pyx
index 3859314505..4983e6e469 100644
--- a/dpctl/program/_program.pyx
+++ b/dpctl/program/_program.pyx
@@ -17,6 +17,7 @@
 # distutils: language = c++
 # cython: language_level=3
 # cython: linetrace=True
+# cython: freethreading_compatible = True
 
 """Implements a Python interface for SYCL's program and kernel runtime classes.
 
diff --git a/dpctl/tests/_cython_api.pyx b/dpctl/tests/_cython_api.pyx
index e694ecb9c3..84e981e040 100644
--- a/dpctl/tests/_cython_api.pyx
+++ b/dpctl/tests/_cython_api.pyx
@@ -16,6 +16,7 @@
 
 # cython: language=c++
 # cython: language_level=3
+# cython: freethreading_compatible = True
 
 cimport dpctl as c_dpctl
 
diff --git a/dpctl/tests/test_sycl_usm.py b/dpctl/tests/test_sycl_usm.py
index 7b79738231..1840ff2a8d 100644
--- a/dpctl/tests/test_sycl_usm.py
+++ b/dpctl/tests/test_sycl_usm.py
@@ -59,7 +59,7 @@ def test_memory_create(memory_ctor):
     assert mobj.sycl_queue == queue
     assert type(repr(mobj)) is str
     assert type(bytes(mobj)) is bytes
-    assert sys.getsizeof(mobj) > nbytes
+    assert sys.getsizeof(mobj) >= nbytes
 
 
 def test_memory_create_with_np():
diff --git a/dpctl/utils/_onetrace_context.py b/dpctl/utils/_onetrace_context.py
index e35269ae91..bdc93d7dcd 100644
--- a/dpctl/utils/_onetrace_context.py
+++ b/dpctl/utils/_onetrace_context.py
@@ -14,6 +14,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import threading
 from contextlib import contextmanager
 from os import environ, getenv
 from platform import system as sys_platform
@@ -25,6 +26,7 @@
 
 _UNCHECKED = sys_platform() == "Linux"
 del sys_platform
+_unchecked_lock = threading.Lock()
 
 
 @contextmanager
@@ -68,21 +70,25 @@ def onetrace_enabled():
     """
     global _UNCHECKED
 
-    if _UNCHECKED:
-        _UNCHECKED = False
-        if not (
-            getenv("PTI_ENABLE", None) == "1"
-            and "onetrace_tool" in getenv("LD_PRELOAD", "")
-        ):
-            import warnings
-
-            warnings.warn(
-                "It looks like Python interpreter was not started using "
-                "`onetrace` utility. Using `onetrace_enabled` may have "
-                "no effect. See `onetrace_enabled.__doc__` for usage.",
-                RuntimeWarning,
-                stacklevel=2,
+    with _unchecked_lock:
+        if _UNCHECKED:
+            _UNCHECKED = False
+            needs_warning = not (
+                getenv("PTI_ENABLE", None) == "1"
+                and "onetrace_tool" in getenv("LD_PRELOAD", "")
             )
+        else:
+            needs_warning = False
+    if needs_warning:
+        import warnings
+
+        warnings.warn(
+            "It looks like Python interpreter was not started using "
+            "`onetrace` utility. Using `onetrace_enabled` may have "
+            "no effect. See `onetrace_enabled.__doc__` for usage.",
+            RuntimeWarning,
+            stacklevel=2,
+        )
 
     _env_var_name = "PTI_ENABLE_COLLECTION"
     saved = getenv(_env_var_name, None)
diff --git a/dpctl/utils/_order_manager.py b/dpctl/utils/_order_manager.py
index fbe2c47763..d0ccf25985 100644
--- a/dpctl/utils/_order_manager.py
+++ b/dpctl/utils/_order_manager.py
@@ -16,7 +16,7 @@ class _SequentialOrderManager:
     def __init__(self):
         self._state = _OrderManager(16)
 
-    def __dealloc__(self):
+    def __del__(self):
         _local = self._state
         SyclEvent.wait_for(_local.get_submitted_events())
         SyclEvent.wait_for(_local.get_host_task_events())
@@ -71,24 +71,31 @@ class SyclQueueToOrderManagerMap:
     def __init__(self):
         self._map = ContextVar(
             "global_order_manager_map",
-            default=defaultdict(_SequentialOrderManager),
+            # no default to avoid sharing a single defaultdict
+            # across threads
         )
 
+    def _get_map(self):
+        """
+        Factory method to get or create a default device queue cache for the
+        current context
+        """
+        try:
+            return self._map.get()
+        except LookupError:
+            m = defaultdict(_SequentialOrderManager)
+            self._map.set(m)
+            return m
+
     def __getitem__(self, q: SyclQueue) -> _SequentialOrderManager:
         """Get order manager for given SyclQueue"""
-        _local = self._map.get()
         if not isinstance(q, SyclQueue):
             raise TypeError(f"Expected `dpctl.SyclQueue`, got {type(q)}")
-        if q in _local:
-            return _local[q]
-        else:
-            v = _local[q]
-            _local[q] = v
-            return v
+        return self._get_map()[q]
 
     def clear(self):
         """Clear content of internal dictionary"""
-        _local = self._map.get()
+        _local = self._get_map()
         for v in _local.values():
             v.wait()
         _local.clear()
diff --git a/dpctl/utils/src/device_queries.cpp b/dpctl/utils/src/device_queries.cpp
index c350a1f280..05fea90e2a 100644
--- a/dpctl/utils/src/device_queries.cpp
+++ b/dpctl/utils/src/device_queries.cpp
@@ -132,7 +132,7 @@ std::uint32_t py_intel_memory_bus_width(const sycl::device &d)
 
 }; // namespace
 
-PYBIND11_MODULE(_device_queries, m)
+PYBIND11_MODULE(_device_queries, m, py::mod_gil_not_used())
 {
     m.def("intel_device_info_device_id", &py_intel_device_id,
           "Get ext_intel_device_id for the device, zero if not an intel device",
diff --git a/dpctl/utils/src/order_keeper.cpp b/dpctl/utils/src/order_keeper.cpp
index 62fc7e7e65..7f0074f91a 100644
--- a/dpctl/utils/src/order_keeper.cpp
+++ b/dpctl/utils/src/order_keeper.cpp
@@ -6,7 +6,7 @@
 #include "sequential_order_keeper.hpp"
 #include <sycl/sycl.hpp>
 
-PYBIND11_MODULE(_seq_order_keeper, m)
+PYBIND11_MODULE(_seq_order_keeper, m, py::mod_gil_not_used())
 {
     py::class_<SequentialOrder>(m, "_OrderManager")
         .def(py::init<std::size_t>())
diff --git a/dpctl/utils/src/sequential_order_keeper.hpp b/dpctl/utils/src/sequential_order_keeper.hpp
index a6cba662b0..f8c6feb107 100644
--- a/dpctl/utils/src/sequential_order_keeper.hpp
+++ b/dpctl/utils/src/sequential_order_keeper.hpp
@@ -3,6 +3,7 @@
 
 #include <algorithm>
 #include <cstddef>
+#include <mutex>
 #include <vector>
 
 namespace
@@ -21,10 +22,12 @@ inline bool is_event_complete(const sycl::event &e)
 class SequentialOrder
 {
 private:
+    mutable std::mutex mu_events;
     std::vector<sycl::event> host_task_events;
     std::vector<sycl::event> submitted_events;
 
-    void prune_complete()
+    // only called with mu_events held
+    void prune_complete_nolock()
     {
         const auto &ht_it =
             std::remove_if(host_task_events.begin(), host_task_events.end(),
@@ -46,76 +49,78 @@ class SequentialOrder
     }
 
     SequentialOrder(const SequentialOrder &other)
-        : host_task_events(other.host_task_events),
-          submitted_events(other.submitted_events)
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(other.mu_events);
+        host_task_events = other.host_task_events;
+        submitted_events = other.submitted_events;
+        prune_complete_nolock();
     }
     SequentialOrder(SequentialOrder &&other)
         : host_task_events{}, submitted_events{}
     {
+        std::lock_guard<std::mutex> lock(other.mu_events);
         host_task_events = std::move(other.host_task_events);
         submitted_events = std::move(other.submitted_events);
-        prune_complete();
+        prune_complete_nolock();
     }
 
     SequentialOrder &operator=(const SequentialOrder &other)
     {
-        host_task_events = other.host_task_events;
-        submitted_events = other.submitted_events;
-
-        prune_complete();
+        if (this != &other) {
+            std::scoped_lock lock(mu_events, other.mu_events);
+            host_task_events = other.host_task_events;
+            submitted_events = other.submitted_events;
+            prune_complete_nolock();
+        }
         return *this;
     }
 
     SequentialOrder &operator=(SequentialOrder &&other)
     {
         if (this != &other) {
+            std::scoped_lock lock(mu_events, other.mu_events);
             host_task_events = std::move(other.host_task_events);
             submitted_events = std::move(other.submitted_events);
-            prune_complete();
+            prune_complete_nolock();
         }
         return *this;
     }
 
     std::size_t get_num_submitted_events() const
     {
+        std::lock_guard<std::mutex> lock(mu_events);
         return submitted_events.size();
     }
 
-    const std::vector<sycl::event> &get_host_task_events()
+    // returns a copy to avoid returning a reference that
+    // could be modified after the lock is released
+    std::vector<sycl::event> get_host_task_events()
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         return host_task_events;
     }
 
-    /*
-        const std::vector<sycl::event> & get_host_task_events() const {
-            return host_task_events;
-        }
-    */
-
     std::size_t get_num_host_task_events() const
     {
+        std::lock_guard<std::mutex> lock(mu_events);
         return host_task_events.size();
     }
 
-    const std::vector<sycl::event> &get_submitted_events()
+    // returns a copy to avoid returning a reference that
+    // could be modified after the lock is released
+    std::vector<sycl::event> get_submitted_events()
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         return submitted_events;
     }
 
-    /*
-        const std::vector<sycl::event> & get_submitted_events() const {
-            return submitted_events;
-        }
-    */
-
     void add_to_both_events(const sycl::event &ht_ev,
                             const sycl::event &comp_ev)
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         if (!is_event_complete(ht_ev))
             host_task_events.push_back(ht_ev);
         if (!is_event_complete(comp_ev))
@@ -125,7 +130,8 @@ class SequentialOrder
     void add_vector_to_both_events(const std::vector<sycl::event> &ht_evs,
                                    const std::vector<sycl::event> &comp_evs)
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         for (const auto &e : ht_evs) {
             if (!is_event_complete(e))
                 host_task_events.push_back(e);
@@ -138,7 +144,8 @@ class SequentialOrder
 
     void add_to_host_task_events(const sycl::event &ht_ev)
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         if (!is_event_complete(ht_ev)) {
             host_task_events.push_back(ht_ev);
         }
@@ -146,7 +153,8 @@ class SequentialOrder
 
     void add_to_submitted_events(const sycl::event &comp_ev)
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         if (!is_event_complete(comp_ev)) {
             submitted_events.push_back(comp_ev);
         }
@@ -155,7 +163,8 @@ class SequentialOrder
     template <std::size_t num>
     void add_list_to_host_task_events(const sycl::event (&ht_events)[num])
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         for (std::size_t i = 0; i < num; ++i) {
             const auto &e = ht_events[i];
             if (!is_event_complete(e))
@@ -166,7 +175,8 @@ class SequentialOrder
     template <std::size_t num>
     void add_list_to_submitted_events(const sycl::event (&comp_events)[num])
     {
-        prune_complete();
+        std::lock_guard<std::mutex> lock(mu_events);
+        prune_complete_nolock();
         for (std::size_t i = 0; i < num; ++i) {
             const auto &e = comp_events[i];
             if (!is_event_complete(e))
@@ -176,8 +186,20 @@ class SequentialOrder
 
     void wait()
     {
-        sycl::event::wait(submitted_events);
-        sycl::event::wait(host_task_events);
-        prune_complete();
+        // snapeshot events outside of mutex to avoid
+        // calling wait inside mutex
+        std::vector<sycl::event> sub_copy;
+        std::vector<sycl::event> ht_copy;
+        {
+            std::lock_guard<std::mutex> lock(mu_events);
+            sub_copy = submitted_events;
+            ht_copy = host_task_events;
+        }
+        sycl::event::wait(sub_copy);
+        sycl::event::wait(ht_copy);
+        {
+            std::lock_guard<std::mutex> lock(mu_events);
+            prune_complete_nolock();
+        }
     }
 };
diff --git a/pyproject.toml b/pyproject.toml
index 92be4c0771..1b37a71ae5 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -29,6 +29,7 @@ classifiers = [
   "Programming Language :: Python :: 3.13",
   "Programming Language :: Python :: 3.14",
   "Programming Language :: Python :: Implementation :: CPython",
+  "Programming Language :: Python :: Free Threading :: 2 - Beta",
   "Topic :: Software Development",
   "Topic :: Scientific/Engineering",
   "Operating System :: Microsoft :: Windows",
@@ -60,7 +61,7 @@ readme = {file = "README.md", content-type = "text/markdown"}
 requires-python = ">=3.10"
 
 [project.optional-dependencies]
-coverage = ["Cython", "pytest", "pytest-cov", "coverage", "tomli"]
+coverage = ["Cython", "pytest", "coverage", "tomli"]
 docs = [
   "Cython",
   "graphviz",