diff --git a/.github/workflows/Benchmark.yml b/.github/workflows/Benchmark.yml
new file mode 100644
index 0000000..36e2d61
--- /dev/null
+++ b/.github/workflows/Benchmark.yml
@@ -0,0 +1,30 @@
+name: Run benchmarks
+on:
+  pull_request:
+    types: [labeled, opened, synchronize, reopened]
+# Only trigger the benchmark job when you add `run benchmark` label to the PR
+jobs:
+  Benchmark:
+    runs-on: ubuntu-latest
+    if: contains(github.event.pull_request.labels.*.name, 'run benchmark')
+    steps:
+      - uses: actions/checkout@v2
+      - uses: julia-actions/setup-julia@latest
+      - name: Cache artifacts
+        uses: actions/cache@v1
+        env:
+          cache-name: cache-artifacts
+        with:
+          path: ~/.julia/artifacts 
+          key: ${{ runner.os }}-test-${{ env.cache-name }}-${{ hashFiles('**/Project.toml') }}
+          restore-keys: |
+            ${{ runner.os }}-test-${{ env.cache-name }}-
+            ${{ runner.os }}-test-
+            ${{ runner.os }}-
+      - name: Install dependencies
+        run: julia -e 'using Pkg; pkg"add JSON PkgBenchmark BenchmarkCI@0.1"'
+      - name: Run benchmarks
+        run: julia benchmark/run_benchmarks.jl
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          
\ No newline at end of file
diff --git a/.gitignore b/.gitignore
index b0c4baa..6ed2db7 100644
--- a/.gitignore
+++ b/.gitignore
@@ -4,6 +4,8 @@
 Manifest.toml
 /docs/build/
 /docs/src/generated/
+/.benchmarkci
+/benchmark/*.json
 
 .vscode
 *.DS_Store
\ No newline at end of file
diff --git a/benchmark/Project.toml b/benchmark/Project.toml
new file mode 100644
index 0000000..f61ac24
--- /dev/null
+++ b/benchmark/Project.toml
@@ -0,0 +1,6 @@
+[deps]
+BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
+ImageCore = "a09fc81d-aa75-5fe9-8630-4744c3626534"
+ImageMagick = "6218d12a-5da1-5696-b52f-db25d2ecc6d1"
+PkgBenchmark = "32113eaa-f34f-5b0d-bd6c-c81e245fc73d"
+TestImages = "5e47fb64-e119-507b-a336-dd2b206d9990"
diff --git a/benchmark/benchmarks.jl b/benchmark/benchmarks.jl
new file mode 100644
index 0000000..fdf53e1
--- /dev/null
+++ b/benchmark/benchmarks.jl
@@ -0,0 +1,48 @@
+using BenchmarkTools
+using DitherPunk
+using TestImages
+using ImageCore
+
+on_CI = haskey(ENV, "GITHUB_ACTIONS")
+
+img_gray = testimage("fabio_gray_256")
+img_color = testimage("fabio_color_256")
+
+## Define color scheme
+white = RGB{Float32}(1, 1, 1)
+yellow = RGB{Float32}(1, 1, 0)
+green = RGB{Float32}(0, 0.5, 0)
+orange = RGB{Float32}(1, 0.5, 0)
+red = RGB{Float32}(1, 0, 0)
+blue = RGB{Float32}(0, 0, 1)
+cs = [white, yellow, green, orange, red, blue]
+
+# Use one representative algorithm of each type
+algs = Dict(
+    "error diffusion" => FloydSteinberg(),
+    "ordered dithering" => Bayer(; level=3),
+    "closest color" => ClosestColor(),
+    "threshold dithering" => WhiteNoiseThreshold(),
+)
+
+# Tag what is tested on each algorithm type
+SUITE = BenchmarkGroup()
+SUITE["error diffusion"] = BenchmarkGroup(["binary", "color"])
+SUITE["ordered dithering"] = BenchmarkGroup(["binary"])
+SUITE["threshold dithering"] = BenchmarkGroup(["binary"])
+SUITE["closest color"] = BenchmarkGroup(["binary", "color"])
+
+println(SUITE)
+
+for (algname, alg) in algs
+    SUITE[algname]["binary new"] = @benchmarkable dither($(img_gray), $(alg))
+    SUITE[algname]["binary inplace"] = @benchmarkable dither!($(copy(img_gray)), $(alg))
+
+    SUITE[algname]["per-channel new"] = @benchmarkable dither($(img_color), SeparateSpace($(alg)))
+    SUITE[algname]["per-channel inplace"] = @benchmarkable dither!($(copy(img_color)), SeparateSpace($(alg)))
+
+    if "color" in SUITE[algname].tags
+        SUITE[algname]["color new"] = @benchmarkable dither($(img_color), $(alg), $(cs))
+        SUITE[algname]["color inplace"] = @benchmarkable dither!($(copy(img_color)), $(alg), $(cs))
+    end
+end
diff --git a/benchmark/run_benchmarks.jl b/benchmark/run_benchmarks.jl
new file mode 100644
index 0000000..6b65413
--- /dev/null
+++ b/benchmark/run_benchmarks.jl
@@ -0,0 +1,10 @@
+# To run benchmarks locally, BenchmarkCI should be added to root project.
+# Then call:
+# ```bash
+# julia benchmark/run_benchmarks.jl
+# ```
+using BenchmarkCI
+on_CI = haskey(ENV, "GITHUB_ACTIONS")
+
+BenchmarkCI.judge()
+on_CI ? BenchmarkCI.postjudge() : BenchmarkCI.displayjudgement()