diff --git a/README.md b/README.md
index d50b9d46..50d9c2ec 100644
--- a/README.md
+++ b/README.md
@@ -114,6 +114,7 @@ extensionCodec | ExtensionCodec | `ExtensinCodec.defaultCodec`
 maxDepth | number | `100`
 initialBufferSize | number | `2048`
 sortKeys | boolean | false
+forceFloat32 | boolean | false
 
 ### `decode(buffer: ArrayLike<number> | ArrayBuffer, options?: DecodeOptions): unknown`
 
diff --git a/src/Encoder.ts b/src/Encoder.ts
index 3a61aa6e..e95b5e0c 100644
--- a/src/Encoder.ts
+++ b/src/Encoder.ts
@@ -18,6 +18,7 @@ export class Encoder {
     readonly maxDepth = DEFAULT_MAX_DEPTH,
     readonly initialBufferSize = DEFAULT_INITIAL_BUFFER_SIZE,
     readonly sortKeys = false,
+    readonly forceFloat32 = false,
   ) {}
 
   encode(object: unknown, depth: number): void {
@@ -118,8 +119,16 @@ export class Encoder {
         }
       }
     } else {
-      this.writeU8(0xcb);
-      this.writeF64(object);
+      // non-integer numbers
+      if (this.forceFloat32) {
+        // float 32
+        this.writeU8(0xca);
+        this.writeF32(object);
+      } else {
+        // float 64
+        this.writeU8(0xcb);
+        this.writeF64(object);
+      }
     }
   }
 
@@ -345,9 +354,14 @@ export class Encoder {
     this.pos += 4;
   }
 
+  writeF32(value: number) {
+    this.ensureBufferSizeToWrite(4);
+    this.view.setFloat32(this.pos, value);
+    this.pos += 4;
+  }
+
   writeF64(value: number) {
     this.ensureBufferSizeToWrite(8);
-
     this.view.setFloat64(this.pos, value);
     this.pos += 8;
   }
diff --git a/src/encode.ts b/src/encode.ts
index cb066b51..94af5307 100644
--- a/src/encode.ts
+++ b/src/encode.ts
@@ -7,6 +7,13 @@ export type EncodeOptions = Partial<
     maxDepth: number;
     initialBufferSize: number;
     sortKeys: boolean;
+
+    /**
+     * If `true`, non-integer numbers are encoded in float32, not in float64 (the default).
+     *
+     * Only use it if precisions don't matter.
+     */
+    forceFloat32: boolean;
   }>
 >;
 
@@ -19,7 +26,13 @@ const defaultEncodeOptions = {};
  * The returned buffer is a slice of a larger `ArrayBuffer`, so you have to use its `#byteOffset` and `#byteLength` in order to convert it to another typed arrays including NodeJS `Buffer`.
  */
 export function encode(value: unknown, options: EncodeOptions = defaultEncodeOptions): Uint8Array {
-  const encoder = new Encoder(options.extensionCodec, options.maxDepth, options.initialBufferSize, options.sortKeys);
+  const encoder = new Encoder(
+    options.extensionCodec,
+    options.maxDepth,
+    options.initialBufferSize,
+    options.sortKeys,
+    options.forceFloat32,
+  );
   encoder.encode(value, 1);
   return encoder.getUint8Array();
 }
diff --git a/test/encode.test.ts b/test/encode.test.ts
index f07d63b3..99aee64b 100644
--- a/test/encode.test.ts
+++ b/test/encode.test.ts
@@ -3,11 +3,28 @@ import { encode, decode } from "@msgpack/msgpack";
 
 describe("encode", () => {
   context("sortKeys", () => {
-    it("canonicalize encoded binaries", () => {
+    it("canonicalizes encoded binaries", () => {
       assert.deepStrictEqual(encode({ a: 1, b: 2 }, { sortKeys: true }), encode({ b: 2, a: 1 }, { sortKeys: true }));
     });
   });
 
+  context("forceFloat32", () => {
+    it("encodes numbers in float64 wihout forceFloat32", () => {
+      assert.deepStrictEqual(encode(3.14), Uint8Array.from([0xcb, 0x40, 0x9, 0x1e, 0xb8, 0x51, 0xeb, 0x85, 0x1f]));
+    });
+
+    it("encodes numbers in float32 when forceFloate32=true", () => {
+      assert.deepStrictEqual(encode(3.14, { forceFloat32: true }), Uint8Array.from([0xca, 0x40, 0x48, 0xf5, 0xc3]));
+    });
+
+    it("encodes numbers in float64 with forceFloat32=false", () => {
+      assert.deepStrictEqual(
+        encode(3.14, { forceFloat32: false }),
+        Uint8Array.from([0xcb, 0x40, 0x9, 0x1e, 0xb8, 0x51, 0xeb, 0x85, 0x1f]),
+      );
+    });
+  });
+
   context("ArrayBuffer as buffer", () => {
     const buffer = encode([1, 2, 3]);
     const arrayBuffer = buffer.buffer.slice(buffer.byteOffset, buffer.byteLength);