diff --git a/.gitignore b/.gitignore
index cf46c7c..da3705c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -2,4 +2,5 @@
 /build
 /Testing
 /cmake-build-debug
+/build-bench
 .vscode
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 9e1149c..ce3c918 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -9,4 +9,11 @@ set(CMAKE_CXX_EXTENSIONS OFF)
 add_subdirectory(src)
 
 include(CTest)
-add_subdirectory(tests)
\ No newline at end of file
+add_subdirectory(tests)
+
+# Benchmarks are disabled by default
+option(BIGINT_BUILD_BENCHMARKS "Build BigInt benchmarks" ON)
+
+if(BIGINT_BUILD_BENCHMARKS)
+    add_subdirectory(benchmarks)
+endif()
\ No newline at end of file
diff --git a/benchmarks/CMakeLists.txt b/benchmarks/CMakeLists.txt
new file mode 100644
index 0000000..7464398
--- /dev/null
+++ b/benchmarks/CMakeLists.txt
@@ -0,0 +1,26 @@
+include(FetchContent)
+
+# Disable Google Benchmark's own tests
+set(BENCHMARK_ENABLE_TESTING OFF CACHE BOOL "" FORCE)
+set(BENCHMARK_ENABLE_GTEST_TESTS OFF CACHE BOOL "" FORCE)
+set(BENCHMARK_ENABLE_INSTALL OFF CACHE BOOL "" FORCE)
+
+FetchContent_Declare(
+        googlebenchmark
+        GIT_REPOSITORY https://github.com/google/benchmark.git
+        GIT_TAG v1.9.5
+        GIT_SHALLOW TRUE
+)
+
+FetchContent_MakeAvailable(googlebenchmark)
+
+add_executable(BigInt-Benchmarks
+        benchmark_bigint.cpp
+)
+
+target_link_libraries(
+        BigInt-Benchmarks
+        PRIVATE
+        BigInt
+        benchmark::benchmark_main
+)
\ No newline at end of file
diff --git a/benchmarks/benchmark_bigint.cpp b/benchmarks/benchmark_bigint.cpp
new file mode 100644
index 0000000..dbeb989
--- /dev/null
+++ b/benchmarks/benchmark_bigint.cpp
@@ -0,0 +1,184 @@
+#include <benchmark/benchmark.h>
+
+#include "bigint.h"
+#include "math_utils.h"
+
+#include <cstddef>
+#include <cstdint>
+
+using operations::BigInt;
+
+namespace {
+
+// Erstellt einen BigInt mit ungefähr/exakt der gewünschten Limb-Größe.
+// Da ein Limb uint64_t ist:
+// 2048 Bit = 32 Limbs.
+BigInt makeBigInt(const std::size_t bits, const std::uint64_t pattern) {
+    const std::size_t limbCount = (bits + 63) / 64;
+
+    ByteArray limbs(limbCount, pattern);
+
+    return BigInt(std::move(limbs));
+}
+
+
+// ------------------------------------------------------------
+// Addition
+// ------------------------------------------------------------
+
+static void BM_Add(benchmark::State& state) {
+    const auto bits = static_cast<std::size_t>(state.range(0));
+
+    const BigInt a = makeBigInt(bits, 0xAAAAAAAAAAAAAAAAULL);
+    const BigInt b = makeBigInt(bits, 0x5555555555555555ULL);
+
+    for (auto _ : state) {
+        BigInt result = a + b;
+
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetComplexityN(static_cast<benchmark::IterationCount>(bits / 64));
+}
+
+BENCHMARK(BM_Add)
+    ->RangeMultiplier(2)
+    ->Range(64, 8192)
+    ->Complexity();
+
+
+// ------------------------------------------------------------
+// Subtraction
+// ------------------------------------------------------------
+
+static void BM_Subtract(benchmark::State& state) {
+    const auto bits = static_cast<std::size_t>(state.range(0));
+
+    // a > b, wichtig bei unsigned BigInt
+    const BigInt a = makeBigInt(bits, 0xFFFFFFFFFFFFFFFFULL);
+    const BigInt b = makeBigInt(bits, 0x5555555555555555ULL);
+
+    for (auto _ : state) {
+        BigInt result = a - b;
+
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetComplexityN(static_cast<benchmark::IterationCount>(bits / 64));
+}
+
+BENCHMARK(BM_Subtract)
+    ->RangeMultiplier(2)
+    ->Range(64, 8192)
+    ->Complexity();
+
+
+// ------------------------------------------------------------
+// Multiplikation
+// ------------------------------------------------------------
+
+static void BM_Multiply(benchmark::State& state) {
+    const auto bits = static_cast<std::size_t>(state.range(0));
+
+    const BigInt a = makeBigInt(bits, 0xAAAAAAAAAAAAAAAAULL);
+    const BigInt b = makeBigInt(bits, 0x5555555555555555ULL);
+
+    for (auto _ : state) {
+        BigInt result = a * b;
+
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetComplexityN(static_cast<benchmark::IterationCount>(bits / 64));
+}
+
+BENCHMARK(BM_Multiply)
+    ->RangeMultiplier(2)
+    ->Range(64, 8192)
+    ->Complexity();
+
+
+// ------------------------------------------------------------
+// Division
+// ------------------------------------------------------------
+
+static void BM_Divide(benchmark::State& state) {
+    const auto bits = static_cast<std::size_t>(state.range(0));
+
+    // a > b
+    const BigInt a = makeBigInt(bits, 0xFFFFFFFFFFFFFFFFULL);
+    const BigInt b = makeBigInt(bits, 0x5555555555555555ULL);
+
+    for (auto _ : state) {
+        BigInt result = a / b;
+
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetComplexityN(static_cast<benchmark::IterationCount>(bits / 64));
+}
+
+BENCHMARK(BM_Divide)
+    ->RangeMultiplier(2)
+    ->Range(64, 4096)
+    ->Complexity();
+
+
+// ------------------------------------------------------------
+// Modulo
+// ------------------------------------------------------------
+
+static void BM_Modulo(benchmark::State& state) {
+    const auto bits = static_cast<std::size_t>(state.range(0));
+
+    const BigInt a = makeBigInt(bits, 0xFFFFFFFFFFFFFFFFULL);
+    const BigInt b = makeBigInt(bits, 0x5555555555555555ULL);
+
+    for (auto _ : state) {
+        BigInt result = a % b;
+
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetComplexityN(static_cast<benchmark::IterationCount>(bits / 64));
+}
+
+BENCHMARK(BM_Modulo)
+    ->RangeMultiplier(2)
+    ->Range(64, 4096)
+    ->Complexity();
+
+
+// ------------------------------------------------------------
+// Modular Exponentiation
+//
+// Verwendet exponent = 65537.
+// Das ist für einen ersten Benchmark wesentlich sinnvoller als
+// direkt einen mehrere tausend Bit großen Exponenten zu verwenden.
+// ------------------------------------------------------------
+
+static void BM_ModPow65537(benchmark::State& state) {
+    const auto bits = static_cast<std::size_t>(state.range(0));
+
+    const BigInt base =
+        makeBigInt(bits, 0x5555555555555555ULL);
+
+    const BigInt exponent(65537);
+
+    // modulus > base
+    const BigInt modulus =
+        makeBigInt(bits, 0xFFFFFFFFFFFFFFFFULL);
+
+    for (auto _ : state) {
+        BigInt result =
+            operations::math::modPow(base, exponent, modulus);
+
+        benchmark::DoNotOptimize(result);
+    }
+}
+
+BENCHMARK(BM_ModPow65537)
+    ->RangeMultiplier(2)
+    ->Range(256, 2048);
+
+} // namespace
\ No newline at end of file
diff --git a/src/bigint.cpp b/src/bigint.cpp
index 0ad2b73..974e6d9 100644
--- a/src/bigint.cpp
+++ b/src/bigint.cpp
@@ -133,58 +133,58 @@ void operations::BigInt::mul(const ByteArray &b) noexcept {
 
     for (size_t i = 0; i < size_a; ++i) {
         std::uint64_t carry = 0;
-        std::uint64_t valA = data[i];
+        const std::uint64_t valA = data[i];
 
-        for (size_t j = 0; j < size_b; ++j) {
-            std::uint64_t valB = b[j];
+        const std::uint64_t n1 = valA >> 32;
+        const std::uint64_t n2 = valA & 0xFFFFFFFFULL;
 
-            std::uint64_t n1 = valA >> 32;
-            std::uint64_t n2 = valA & 0xFFFFFFFFULL;
+        for (size_t j = 0; j < size_b; ++j) {
+            const std::uint64_t valB = b[j];
 
-            std::uint64_t n3 = valB >> 32;
-            std::uint64_t n4 = valB & 0xFFFFFFFFULL;
+            const std::uint64_t n3 = valB >> 32;
+            const std::uint64_t n4 = valB & 0xFFFFFFFFULL;
 
-            std::uint64_t r1 = n1 * n3;
+            const std::uint64_t r1 = n1 * n3;
             std::uint64_t r2 = n2 * n4;
-            std::uint64_t r3 = n1 * n4;
-            std::uint64_t r4 = n2 * n3;
+            const std::uint64_t r3 = n1 * n4;
+            const std::uint64_t r4 = n2 * n3;
 
-            std::uint64_t c1 = r1 >> 32;
-            std::uint64_t c2 = r3 >> 32;
-            std::uint64_t fr1 = (r1 << 32) | (r3 & 0xFFFFFFFFULL);
-            std::uint64_t fc1 = (c1 << 32) | c2;
+            const std::uint64_t c1 = r1 >> 32;
+            const std::uint64_t c2 = r3 >> 32;
+            const std::uint64_t fr1 = (r1 << 32) | (r3 & 0xFFFFFFFFULL);
+            const std::uint64_t fc1 = (c1 << 32) | c2;
 
-            std::uint64_t c3 = r2 >> 32;
+            const std::uint64_t c3 = r2 >> 32;
             r2 &= 0xFFFFFFFFULL;
-            std::uint64_t c4 = r4 >> 32;
-            std::uint64_t fr2 = (r4 << 32) | r2;
-            std::uint64_t fc2 = c3 | (c4 << 32);
+            const std::uint64_t c4 = r4 >> 32;
+            const std::uint64_t fr2 = (r4 << 32) | r2;
+            const std::uint64_t fc2 = c3 | (c4 << 32);
 
             std::uint64_t fl = fr2;
             std::uint64_t runningCarry = 0;
-            std::uint64_t valFL = fc2 << 32;
+            const std::uint64_t valFL = fc2 << 32;
             fl += valFL;
             if (valFL > fl) {
                 runningCarry += 1ULL;
             }
 
-            std::uint64_t valFH = fr1 << 32;
+            const std::uint64_t valFH = fr1 << 32;
             fl += valFH;
             if (valFH > fl) {
                 runningCarry += 1ULL;
             }
 
-            std::uint64_t fh = (fc2 >> 32) + (fr1 >> 32) + fc1 + runningCarry;
+            const std::uint64_t fh = (fc2 >> 32) + (fr1 >> 32) + fc1 + runningCarry;
 
 
-            std::uint64_t current = result[i + j];
+            const std::uint64_t current = result[i + j];
 
             std::uint64_t sum_low = current + fl;
-            std::uint64_t carry_out1 = (sum_low < current) ? 1ULL : 0ULL;
+            const std::uint64_t carry_out1 = (sum_low < current) ? 1ULL : 0ULL;
 
-            std::uint64_t sum_low_prev = sum_low;
+            const std::uint64_t sum_low_prev = sum_low;
             sum_low += carry;
-            std::uint64_t carry_out2 = (sum_low < sum_low_prev) ? 1ULL : 0ULL;
+            const std::uint64_t carry_out2 = (sum_low < sum_low_prev) ? 1ULL : 0ULL;
 
             result[i + j] = sum_low;