diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index a7e23b7f18..6496121013 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -371,6 +371,29 @@ jobs:
       - name: Run process replay tests
         uses: ./.github/actions/process-replay
 
+  testmodels:
+    name: Models (llvm+clang+gpu)
+    runs-on: ubuntu-22.04
+    timeout-minutes: 10
+    steps:
+      - name: Checkout Code
+        uses: actions/checkout@v4
+      - name: Setup Environment
+        uses: ./.github/actions/setup-tinygrad
+        with:
+          key: models
+          deps: testing
+          opencl: 'true'
+          llvm: 'true'
+      - name: Test models (llvm)
+        run: LLVM=1 python -m pytest -n=auto test/models --durations=20
+      - name: Test models (gpu)
+        run: GPU=1 python -m pytest -n=auto test/models --durations=20
+      - name: Test models (clang)
+        run: CLANG=1 python -m pytest -n=auto test/models --durations=20
+      - name: Run process replay tests
+        uses: ./.github/actions/process-replay
+
   testdsp:
     name: Linux (DSP)
     runs-on: ubuntu-24.04
@@ -404,8 +427,8 @@ jobs:
     - name: Setup Environment
       uses: ./.github/actions/setup-tinygrad
       with:
-        key: webgpu
-        deps: testing
+        key: webgpu-minimal
+        deps: testing_minimal
         python-version: '3.11'
         webgpu: 'true'
     - name: Check Device.DEFAULT (WEBGPU) and print some source
@@ -414,13 +437,12 @@ jobs:
         WEBGPU=1 DEBUG=4 FORWARD_ONLY=1 python3 test/test_ops.py TestOps.test_add
     - name: Run selected webgpu tests
       run: |
-          WEBGPU=1 python3 -m pytest -n=auto test/ --ignore=test/external --ignore=test/models --ignore=test/unit \
+          WEBGPU=1 python3 -m pytest -n=auto test/ --ignore=test/models --ignore=test/unit \
           --ignore=test/test_copy_speed.py --ignore=test/test_rearrange_einops.py --ignore=test/test_speed_v_torch.py --ignore=test/test_transcendental.py \
           --ignore=test/test_fuzz_shape_ops.py --ignore=test/test_linearizer_failures.py --durations=20
     - name: Run process replay tests
       uses: ./.github/actions/process-replay
 
-
   tests:
     strategy:
       fail-fast: false
@@ -437,8 +459,8 @@ jobs:
       - name: Setup Environment
         uses: ./.github/actions/setup-tinygrad
         with:
-          key: ${{ matrix.backend }}
-          deps: testing${{matrix.backend=='ptx'&&',cuda'||matrix.backend=='triton'&&',triton'||''}}
+          key: ${{ matrix.backend }}-minimal
+          deps: testing_minimal${{matrix.backend=='ptx'&&',cuda'||matrix.backend=='triton'&&',triton'||''}}
           opencl: ${{ matrix.backend == 'gpu' && 'true' }}
           amd: ${{ matrix.backend == 'amd' && 'true' }}
           cuda: ${{ (matrix.backend == 'ptx' || matrix.backend == 'triton' || matrix.backend == 'nv') && 'true' }}
@@ -450,10 +472,10 @@ jobs:
           DEBUG=5 PYTHONPATH=${{ github.workspace }} FORWARD_ONLY=1 python3 test/test_ops.py TestOps.test_add
       - name: Run pytest (not cuda or amd)
         if: matrix.backend!='ptx' && matrix.backend!='triton' && matrix.backend != 'amd' && matrix.backend != 'nv'
-        run: python -m pytest -n=auto test/ --ignore=test/unit --durations=20
+        run: python -m pytest -n=auto test/ --ignore=test/models --ignore=test/unit --durations=20
       - name: Run pytest (cuda)
         if: matrix.backend=='ptx'||matrix.backend=='triton'||matrix.backend=='nv'
-        run: python -m pytest -n=auto test/ -k 'not (half or test_efficientnet_safetensors)' --ignore=test/external --ignore=test/models --ignore=test/unit --ignore test/test_gc.py --durations=20
+        run: python -m pytest -n=auto test/ -k 'not (half or test_efficientnet_safetensors)' --ignore=test/models --ignore=test/unit --ignore test/test_gc.py --durations=20
       - name: Run pytest (amd)
         if: matrix.backend=='amd'
         run: python -m pytest -n=auto test/test_ops.py test/test_dtype.py test/test_dtype_alu.py test/test_linearizer.py test/test_randomness.py test/imported/test_indexing.py test/test_hcq.py test/external/external_test_am.py --durations=20
@@ -478,8 +500,11 @@ jobs:
         key: metal2
         deps: testing
         python-version: '3.11'
+        amd: 'true'
     - name: Run real world test
       run: JIT=2 METAL=1 python -m pytest -n=auto test/models/test_real_world.py --durations=20
+    - name: Test models (Metal)
+      run: JIT=2 METAL=1 python -m pytest -n=auto test/models --durations=20
     - name: Run ONNX
       run: JIT=2 METAL=1 python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
     - name: Test tensor core ops (fake)
@@ -496,29 +521,13 @@ jobs:
     # run: FUZZ_SCHEDULE=1 FUZZ_SCHEDULE_MAX_PATHS=5 python -m pytest test/models/test_train.py test/models/test_end2end.py
     - name: Run TRANSCENDENTAL math
       run: TRANSCENDENTAL=2 python -m pytest -n=auto test/test_ops.py::TestOps::test_sin test/test_ops.py::TestOps::test_cos test/test_ops.py::TestOps::test_tan test/test_ops.py::TestOps::test_exp test/test_ops.py::TestOps::test_log --durations=20
-    - name: Run process replay tests
-      uses: ./.github/actions/process-replay
-
-  testmetal:
-    name: MacOS (metal)
-    runs-on: macos-14
-    timeout-minutes: 10
-
-    steps:
-    - name: Checkout Code
-      uses: actions/checkout@v4
-    - name: Setup Environment
-      uses: ./.github/actions/setup-tinygrad
-      with:
-        key: metal
-        deps: testing
-        python-version: '3.11'
-    - name: Check Device.DEFAULT (METAL) and print some source
+    - name: Run pytest (amd)
+      env:
+        MOCKGPU: 1
+        AMD: 1
+        FORWARD_ONLY: 1
       run: |
-        METAL=1 python -c "from tinygrad import Device; assert Device.DEFAULT == 'METAL', Device.DEFAULT"
-        METAL=1 DEBUG=4 FORWARD_ONLY=1 python3 test/test_ops.py TestOps.test_add
-    - name: Run metal test
-      run: JIT=2 METAL=1 python -m pytest -n=auto test/ --ignore=test/external --ignore=test/models --ignore=test/unit --durations=20
+        python3 -m pytest -n=auto test/test_hcq.py test/test_tiny.py --durations=20
     - name: Run process replay tests
       uses: ./.github/actions/process-replay
 
@@ -543,27 +552,12 @@ jobs:
     - name: Run WEBGPU Efficientnet
       run: node test/web/test_webgpu.js
 
-  osxclang:
-    name: MacOS (clang)
-    runs-on: macos-15
-    timeout-minutes: 10
-    steps:
-      - name: Checkout Code
-        uses: actions/checkout@v4
-      - name: Setup Environment
-        uses: ./.github/actions/setup-tinygrad
-        with:
-          key: macos-clang
-          deps: testing
-      - name: Run pytest (clang)
-        env:
-          CLANG: 1
-        run: |
-          python3 -m pytest -n=auto test/ --ignore=test/unit --durations=20
-          ! (DEBUG=7 python3 test/test_zero_copy.py 2>&1 || true) | grep -E '^0x.*[^0](x18|w18).*$'
-
   osxtests:
-    name: MacOS (amd+llvm)
+    strategy:
+      fail-fast: false
+      matrix:
+        backend: [metal, llvm, clang]
+    name: MacOS (${{ matrix.backend }})
     runs-on: macos-15
     timeout-minutes: 10
     steps:
@@ -572,23 +566,19 @@ jobs:
       - name: Setup Environment
         uses: ./.github/actions/setup-tinygrad
         with:
-          key: macos-other
-          deps: testing
-          amd: 'true'
-          llvm: 'true'
-      - name: Run pytest (amd)
-        env:
-          MOCKGPU: 1
-          AMD: 1
-          FORWARD_ONLY: 1
+          key: macos-${{ matrix.backend }}-minimal
+          deps: testing_minimal
+          llvm: ${{ matrix.backend == 'llvm' && 'true' }}
+      - name: Set env
+        run: printf "${{ matrix.backend == 'llvm' && 'LLVM=1' || matrix.backend == 'clang' && 'CLANG=1' || matrix.backend == 'metal' && 'METAL=1\nJIT=2'}}" >> $GITHUB_ENV
+      - name: Check Device.DEFAULT and print some source
         run: |
-          python3 -m pytest -n=auto test/test_hcq.py test/test_tiny.py --durations=20
-      - name: Run pytest (llvm)
-        env:
-          LLVM: 1
-        run: |
-          python3 -m pytest -n=auto test/ --ignore=test/unit --durations=20
-          ! (DEBUG=7 python3 test/test_zero_copy.py 2>&1 || true) | grep -E '^0x.*[^0](x18|w18).*$'
+          python -c "from tinygrad import Device; assert Device.DEFAULT == '${{ matrix.backend }}'.upper(), Device.DEFAULT"
+          DEBUG=4 python3 test/test_tiny.py TestTiny.test_plus
+      - name: Run pytest (${{ matrix.backend }})
+        run: python3 -m pytest -n=auto test/ --ignore=test/models --ignore=test/unit --durations=20
+      - name: Run process replay tests
+        uses: ./.github/actions/process-replay
 
 # ****** Windows Tests ******