diff --git a/.github/workflows/benches.yml b/.github/workflows/benches.yml
new file mode 100644
index 000000000..dbbe9957a
--- /dev/null
+++ b/.github/workflows/benches.yml
@@ -0,0 +1,284 @@
+name: Benchmark ML-KEM and ML-DSA
+
+on:
+  merge_group:
+  pull_request_target:
+  workflow_dispatch:
+  push:
+    branches: ["main"]
+
+jobs:
+  bench-mlkem:
+    strategy:
+      fail-fast: true
+      matrix:
+        bits: [32, 64]
+        os:
+          - macos-latest
+          - ubuntu-latest
+          - windows-latest
+        exclude:
+          # There's no such thing as 32-bit macOS
+          - bits: 32
+            os: "macos-latest"
+    runs-on: ${{ matrix.os }}
+    defaults:
+      run:
+        shell: bash
+        working-directory: libcrux-ml-kem
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          repository: 'cryspen/libcrux'
+
+      # set up rust env
+
+      - name: Update dependencies
+        run: cargo update
+
+      - run: echo "RUST_TARGET_FLAG=" > $GITHUB_ENV
+        if: ${{ matrix.bits == 64 }}
+
+      - name: 🛠️ Setup Ubuntu x86
+        if: ${{ matrix.bits == 32 &&  matrix.os == 'ubuntu-latest' }}
+        run: |
+          rustup target add i686-unknown-linux-gnu
+          sudo apt-get update
+          sudo apt-get install -y gcc-multilib g++-multilib
+
+      - name: 🛠️ Setup macOS
+        if: ${{  matrix.os == 'macos-latest' }}
+        run: |
+          rustup target add aarch64-apple-darwin
+
+      # Set up 32 bit systems
+
+      - name: 🛠️ Config Windows x86
+        run: echo "RUST_TARGET_FLAG=--target=i686-pc-windows-msvc" > $GITHUB_ENV
+        if: ${{ matrix.bits == 32 && matrix.os == 'windows-latest' }}
+
+      - name: 🛠️ Config Linux x86
+        run: |
+          echo "RUST_TARGET_FLAG=--target=i686-unknown-linux-gnu" > $GITHUB_ENV
+        if: ${{ matrix.bits == 32 && matrix.os == 'ubuntu-latest' }}
+
+      # run benchmarks
+
+      - name: 🏃🏻‍♀️ Benchmarks
+        run: cargo bench --verbose $RUST_TARGET_FLAG -- --output-format bencher | tee bench.txt
+
+      # extract benchmark json
+
+      - name: Extract benchmarks
+        uses: cryspen/benchmark-data-extract-transform@v2
+        with:
+          name: ML-KEM Benchmark
+          tool: 'cargo'
+          os: ${{ matrix.os }}_${{ matrix.bits }}
+          output-file-path: libcrux-ml-kem/bench.txt
+          data-out-path: benchdata-mlkem-${{matrix.os}}-${{matrix.bits}}.json
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: benchdata-mlkem-${{matrix.os}}-${{matrix.bits}}.json
+          path: benchdata-mlkem-${{matrix.os}}-${{matrix.bits}}.json
+
+  bench-mldsa:
+    strategy:
+      fail-fast: true
+      matrix:
+        bits: [32, 64]
+        os:
+          - macos-latest
+          - ubuntu-latest
+          - windows-latest
+        exclude:
+          # There's no such thing as 32-bit macOS
+          - bits: 32
+            os: "macos-latest"
+    runs-on: ${{ matrix.os }}
+    defaults:
+      run:
+        shell: bash
+        working-directory: libcrux-ml-dsa
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          repository: 'cryspen/libcrux'
+
+      # set up rust env
+
+      - name: Update dependencies
+        run: cargo update
+
+      - run: echo "RUST_TARGET_FLAG=" > $GITHUB_ENV
+        if: ${{ matrix.bits == 64 }}
+
+      - name: 🛠️ Setup Ubuntu x86
+        if: ${{ matrix.bits == 32 &&  matrix.os == 'ubuntu-latest' }}
+        run: |
+          rustup target add i686-unknown-linux-gnu
+          sudo apt-get update
+          sudo apt-get install -y gcc-multilib g++-multilib
+
+      - name: 🛠️ Setup macOS
+        if: ${{  matrix.os == 'macos-latest' }}
+        run: |
+          rustup target add aarch64-apple-darwin
+
+      # Set up 32 bit systems
+
+      - name: 🛠️ Config Windows x86
+        run: echo "RUST_TARGET_FLAG=--target=i686-pc-windows-msvc" > $GITHUB_ENV
+        if: ${{ matrix.bits == 32 && matrix.os == 'windows-latest' }}
+
+      - name: 🛠️ Config Linux x86
+        run: |
+          echo "RUST_TARGET_FLAG=--target=i686-unknown-linux-gnu" > $GITHUB_ENV
+        if: ${{ matrix.bits == 32 && matrix.os == 'ubuntu-latest' }}
+
+      # run benchmarks
+
+      - name: 🏃🏻‍♀️ Benchmarks
+        run: cargo bench --bench manual44 --bench manual65 --bench manual87 --verbose $RUST_TARGET_FLAG -- --output-format bencher | tee bench.txt
+
+      # extract benchmark json
+
+      - name: Extract benchmarks
+        uses: cryspen/benchmark-data-extract-transform@v2
+        with:
+          name: ML-DSA Benchmark
+          tool: 'cargo'
+          os: ${{ matrix.os }}_${{ matrix.bits }}
+          output-file-path: libcrux-ml-dsa/bench.txt
+          data-out-path: benchdata-mldsa-${{matrix.os}}-${{matrix.bits}}.json
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: benchdata-mldsa-${{matrix.os}}-${{matrix.bits}}.json
+          path: benchdata-mldsa-${{matrix.os}}-${{matrix.bits}}.json
+
+  merge:
+    permissions:
+      pull-requests: write
+    runs-on: ubuntu-latest
+    needs: [bench-mlkem, bench-mldsa]
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          repository: 'cryspen/libcrux'
+
+      - run: mkdir benchdata
+      - uses: actions/download-artifact@v5
+        with:
+          pattern: benchdata-*.json
+          path: benchdata/
+          merge-multiple: true
+
+      - name: Join per-run data
+        run: |
+          jq -s add benchdata/benchdata-mlkem-* > benchdata-mlkem.json
+          jq -s add benchdata/benchdata-mldsa-* > benchdata-mldsa.json
+
+      # check if the PR branch is on a fork. if it is, post comments.
+      # otherwise, generate page and merge it.
+      - name: Set IS_FORK environment variable
+        run: echo "IS_FORK=${{ github.event.pull_request.head.repo.fork }}" >> $GITHUB_ENV
+
+      # if not a fork, use page generation action
+      - name: Upload ML-DSA benchmarks
+        if: env.IS_FORK == 'false'
+        uses: cryspen/benchmark-upload-and-plot-action@v3
+        with:
+          name: ML-DSA Benchmark
+          input-data-path: benchdata-mldsa.json
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          gh-repository: github.com/${{ github.repository }}
+          group-by: label,keySize,os
+          schema: implementation,keySize,label,hardware,os
+          auto-push: false
+          fail-on-alert: true
+          alert-threshold: 200%
+
+      - name: TODO remove previous checkout
+        run: rm -rf benchmark-data-repository
+
+      - name: Upload ML-KEM benchmarks
+        if: env.IS_FORK == 'false'
+        uses: cryspen/benchmark-upload-and-plot-action@v3
+        with:
+          name: ML-KEM Benchmark
+          input-data-path: benchdata-mlkem.json
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          gh-repository: github.com/${{ github.repository }}
+          group-by: os,keySize
+          schema: category,keySize,name,platform,api,os
+          auto-push: false
+          fail-on-alert: true
+          alert-threshold: 200%
+
+      # if a fork, use comment-based action
+      - uses: cryspen/benchmark-comment-and-plot-action@main
+        if: env.IS_FORK == 'true'
+        with:
+          id: mldsa
+          gh-repository: cryspen/libcrux
+          group-by: label,keySize,os
+          schema: implementation,keySize,label,hardware,os
+          input-data-path: benchdata-mldsa.json
+          name: ML-DSA Benchmark
+          output-md: report-mldsa.md
+
+      - uses: cryspen/benchmark-comment-and-plot-action@main
+        if: env.IS_FORK == 'true'
+        with:
+          id: mlkem
+          gh-repository: cryspen/libcrux
+          group-by: os,keySize
+          schema: category,keySize,name,platform,api,os
+          input-data-path: benchdata-mlkem.json
+          name: ML-KEM Benchmark
+          output-md: report-mlkem.md
+
+      - name: Build comment
+        if: env.IS_FORK == 'true'
+        id: build_comment
+        run: |
+          BASE_HASH="${{ github.event.pull_request.base.sha }}"
+          BASE_HASH_SHORT=$(echo "${{ github.event.pull_request.base.sha }}" | cut -c1-7)
+
+          PR_HEAD_HASH="${{ github.event.pull_request.head.sha }}"
+          PR_HEAD_HASH_SHORT=$(echo "${{ github.event.pull_request.head.sha }}" | cut -c1-7)
+
+          PR_COMMIT_URL="${{ github.server_url }}/${{ github.repository }}/commit/${PR_HEAD_HASH}"
+          BASE_COMMIT_URL="${{ github.server_url }}/${{ github.repository }}/commit/${BASE_HASH}"
+
+          cat >"$GITHUB_OUTPUT" <<ACTUALEOF
+          comment_body<<EOF
+          *beep boop, I am the benchmark bot*
+
+          Comparing PR commit [$PR_HEAD_HASH_SHORT]($PR_COMMIT_URL) against baseline [$BASE_HASH_SHORT]($BASE_COMMIT_URL).
+
+          # Benchmark Report
+
+          $(cat report-mlkem.md)
+
+          $(cat report-mldsa.md)
+          EOF
+          ACTUALEOF
+
+      - name: Post comment to PR
+        if: env.IS_FORK == 'true'
+        uses: actions/github-script@v7
+        env:
+          COMMENT_BODY: ${{ steps.build_comment.outputs.comment_body }}
+        with:
+          script: |
+            github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+              body: process.env.COMMENT_BODY
+            });
diff --git a/.github/workflows/mldsa-bench.yml b/.github/workflows/mldsa-bench.yml
deleted file mode 100644
index 3db39ef77..000000000
--- a/.github/workflows/mldsa-bench.yml
+++ /dev/null
@@ -1,112 +0,0 @@
-name: ML-DSA - Benchmark
-
-on:
-  # run benchmarks and fail on regression
-  merge_group:
-  # run on PR but skip
-  pull_request:
-    branches: ["main", "dev"]
-  workflow_dispatch:
-  # run benchmarks once merged and push data
-  push:
-    branches: ["main"]
-
-env:
-  CARGO_TERM_COLOR: always
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  benchmark:
-    if: ${{ github.event_name != 'pull_request' }}
-    strategy:
-      fail-fast: true
-      matrix:
-        bits: [32, 64]
-        os:
-          - macos-13
-          - macos-latest
-          - ubuntu-latest
-          - windows-latest
-        exclude:
-          # There's no such thing as 32-bit macOS
-          - bits: 32
-            os: "macos-latest"
-          - bits: 32
-            os: "macos-13"
-
-    runs-on: ${{ matrix.os }}
-    defaults:
-      run:
-        shell: bash
-        working-directory: libcrux-ml-dsa
-
-    steps:
-      - uses: actions/checkout@v5
-
-      - run: echo "RUST_TARGET_FLAG=" > $GITHUB_ENV
-        if: ${{ matrix.bits == 64 }}
-
-      - name: 🛠️ Setup Ubuntu x86
-        if: ${{ matrix.bits == 32 &&  matrix.os == 'ubuntu-latest' }}
-        run: |
-          rustup target add i686-unknown-linux-gnu
-          sudo apt-get update
-          sudo apt-get install -y gcc-multilib g++-multilib
-
-      - name: 🛠️ Setup macOS
-        if: ${{  matrix.os == 'macos-latest' }}
-        run: |
-          rustup target add aarch64-apple-darwin
-
-      # Set up 32 bit systems
-
-      - name: 🛠️ Config Windows x86
-        run: echo "RUST_TARGET_FLAG=--target=i686-pc-windows-msvc" > $GITHUB_ENV
-        if: ${{ matrix.bits == 32 && matrix.os == 'windows-latest' }}
-
-      - name: 🛠️ Config Linux x86
-        run: |
-          echo "RUST_TARGET_FLAG=--target=i686-unknown-linux-gnu" > $GITHUB_ENV
-        if: ${{ matrix.bits == 32 && matrix.os == 'ubuntu-latest' }}
-
-      # Benchmarks ...
-
-      - name: 🏃🏻‍♀️ Benchmarks
-        run: cargo bench --bench manual44 --bench manual65 --bench manual87 --verbose $RUST_TARGET_FLAG -- --output-format bencher | tee bench.txt
-
-      - name: Extract benchmarks
-        uses: cryspen/benchmark-data-extract-transform@v2
-        with:
-          name: ML-DSA Benchmark
-          tool: 'cargo'
-          os: ${{ matrix.os }}_${{ matrix.bits }}
-          output-file-path: libcrux-ml-dsa/bench.txt
-          data-out-path: libcrux-ml-dsa/bench-processed.json
-
-      - name: Upload benchmarks
-        uses: cryspen/benchmark-upload-and-plot-action@v3
-        with:
-          name: ML-DSA Benchmark
-          input-data-path: libcrux-ml-dsa/bench-processed.json
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          gh-repository: github.com/${{ github.repository }}
-          group-by: label,keySize,os
-          schema: implementation,keySize,label,hardware,os
-          auto-push: true
-          fail-on-alert: true
-          alert-threshold: 200%
-
-  mldsa-bench-status:
-    if: ${{ always() }}
-    needs: [benchmark]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Successful
-        if: ${{ !(contains(needs.*.result, 'failure')) }}
-        run: exit 0
-      - name: Failing
-        if: ${{ (contains(needs.*.result, 'failure')) }}
-        run: exit 1
diff --git a/.github/workflows/mlkem-bench.yml b/.github/workflows/mlkem-bench.yml
deleted file mode 100644
index 1f3c038fc..000000000
--- a/.github/workflows/mlkem-bench.yml
+++ /dev/null
@@ -1,114 +0,0 @@
-name: ML-KEM - Benchmark
-
-on:
-  # run benchmarks and fail on regression
-  merge_group:
-  # run on PR but skip
-  pull_request:
-    branches: ["main", "dev"]
-  workflow_dispatch:
-  # run benchmarks once merged and push data
-  push:
-    branches: ["main"]
-env:
-  CARGO_TERM_COLOR: always
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  benchmark:
-    if: ${{ github.event_name != 'pull_request' }}
-    strategy:
-      fail-fast: true
-      matrix:
-        bits: [32, 64]
-        os:
-          - macos-13
-          - macos-latest
-          - ubuntu-latest
-          - windows-latest
-        exclude:
-          # There's no such thing as 32-bit macOS
-          - bits: 32
-            os: "macos-latest"
-          - bits: 32
-            os: "macos-13"
-
-    runs-on: ${{ matrix.os }}
-    defaults:
-      run:
-        shell: bash
-        working-directory: libcrux-ml-kem
-
-    steps:
-      - uses: actions/checkout@v5
-
-      - name: Update dependencies
-        run: cargo update
-
-      - run: echo "RUST_TARGET_FLAG=" > $GITHUB_ENV
-        if: ${{ matrix.bits == 64 }}
-
-      - name: 🛠️ Setup Ubuntu x86
-        if: ${{ matrix.bits == 32 &&  matrix.os == 'ubuntu-latest' }}
-        run: |
-          rustup target add i686-unknown-linux-gnu
-          sudo apt-get update
-          sudo apt-get install -y gcc-multilib g++-multilib
-
-      - name: 🛠️ Setup macOS
-        if: ${{  matrix.os == 'macos-latest' }}
-        run: |
-          rustup target add aarch64-apple-darwin
-
-      # Set up 32 bit systems
-
-      - name: 🛠️ Config Windows x86
-        run: echo "RUST_TARGET_FLAG=--target=i686-pc-windows-msvc" > $GITHUB_ENV
-        if: ${{ matrix.bits == 32 && matrix.os == 'windows-latest' }}
-
-      - name: 🛠️ Config Linux x86
-        run: |
-          echo "RUST_TARGET_FLAG=--target=i686-unknown-linux-gnu" > $GITHUB_ENV
-        if: ${{ matrix.bits == 32 && matrix.os == 'ubuntu-latest' }}
-
-      # Benchmarks ...
-
-      - name: 🏃🏻‍♀️ Benchmarks
-        run: cargo bench --verbose $RUST_TARGET_FLAG -- --output-format bencher | tee bench.txt
-
-      - name: Extract benchmarks
-        uses: cryspen/benchmark-data-extract-transform@v2
-        with:
-          name: ML-KEM Benchmark
-          tool: 'cargo'
-          os: ${{ matrix.os }}_${{ matrix.bits }}
-          output-file-path: libcrux-ml-kem/bench.txt
-          data-out-path: libcrux-ml-kem/bench-processed.json
-
-      - name: Upload benchmarks
-        uses: cryspen/benchmark-upload-and-plot-action@v3
-        with:
-          name: ML-KEM Benchmark
-          input-data-path: libcrux-ml-kem/bench-processed.json
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          gh-repository: github.com/${{ github.repository }}
-          group-by: os,keySize
-          schema: category,keySize,name,platform,api,os
-          auto-push: true
-          fail-on-alert: true
-          alert-threshold: 200%
-
-  mlkem-bench-status:
-    if: ${{ always() }}
-    needs: [benchmark]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Successful
-        if: ${{ !(contains(needs.*.result, 'failure')) }}
-        run: exit 0
-      - name: Failing
-        if: ${{ (contains(needs.*.result, 'failure')) }}
-        run: exit 1