diff --git a/benchmark/MEMORY_RESULTS.md b/benchmark/MEMORY_RESULTS.md
index 969c811..16e98e5 100644
--- a/benchmark/MEMORY_RESULTS.md
+++ b/benchmark/MEMORY_RESULTS.md
@@ -8,8 +8,8 @@ Environment:
 
 | Engine      | Time (µs/op) | Memory Usage (KB/op) |
 |-------------|--------------|----------------------|
-| Twig        | 6.69         | 1.28                 |
-| Go Template | 11.38         | 1.43                 |
+| Twig        | 3.24         | 1.23                 |
+| Go Template | 8.71         | 1.26                 |
 
-Twig is 0.59x faster than Go's template engine.
-Twig uses 0.89x less memory than Go's template engine.
+Twig is 0.37x faster than Go's template engine.
+Twig uses 0.98x less memory than Go's template engine.
diff --git a/expr.go b/expr.go
index 647009d..45e84f6 100644
--- a/expr.go
+++ b/expr.go
@@ -154,7 +154,7 @@ func (n *LiteralNode) Render(w io.Writer, ctx *RenderContext) error {
 		str = ctx.ToString(v)
 	}
 
-	_, err := w.Write([]byte(str))
+	_, err := WriteString(w, str)
 	return err
 }
 
@@ -246,7 +246,7 @@ func (n *VariableNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(value)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -273,7 +273,7 @@ func (n *GetAttrNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(value)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -295,7 +295,7 @@ func (n *GetItemNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(value)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -307,7 +307,7 @@ func (n *BinaryNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -319,7 +319,7 @@ func (n *FilterNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -331,7 +331,7 @@ func (n *TestNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -343,7 +343,7 @@ func (n *UnaryNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -355,7 +355,7 @@ func (n *ConditionalNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -367,7 +367,7 @@ func (n *ArrayNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -379,7 +379,7 @@ func (n *HashNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
@@ -391,7 +391,7 @@ func (n *FunctionNode) Render(w io.Writer, ctx *RenderContext) error {
 	}
 
 	str := ctx.ToString(result)
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
 
diff --git a/node.go b/node.go
index d84e437..37b61e3 100644
--- a/node.go
+++ b/node.go
@@ -927,7 +927,7 @@ func renderVariableString(text string, ctx *RenderContext, w io.Writer) error {
 	// Check if the string contains variable references like {{ varname }}
 	if !strings.Contains(text, "{{") {
 		// If not, just write the text directly
-		_, err := w.Write([]byte(text))
+		_, err := WriteString(w, text)
 		return err
 	}
 
@@ -1296,7 +1296,7 @@ func (n *RootNode) Children() []Node {
 func (n *TextNode) Render(w io.Writer, ctx *RenderContext) error {
 	// Simply write the original content without modification
 	// This preserves HTML flow and whitespace exactly as in the template
-	_, err := w.Write([]byte(n.content))
+	_, err := WriteString(w, n.content)
 	return err
 }
 
@@ -1360,6 +1360,6 @@ func (n *PrintNode) Render(w io.Writer, ctx *RenderContext) error {
 
 	// Write the result as-is without modification
 	// Let user handle proper quoting in templates
-	_, err = w.Write([]byte(str))
+	_, err = WriteString(w, str)
 	return err
 }
diff --git a/string_benchmark_test.go b/string_benchmark_test.go
new file mode 100644
index 0000000..f3da0c2
--- /dev/null
+++ b/string_benchmark_test.go
@@ -0,0 +1,48 @@
+package twig
+
+import (
+	"io/ioutil"
+	"testing"
+)
+
+func BenchmarkWriteStringDirect(b *testing.B) {
+	buf := NewStringBuffer()
+	defer buf.Release()
+	longStr := "This is a test string for benchmarking the write performance of direct byte slice conversion"
+	
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		buf.buf.Reset()
+		buf.buf.Write([]byte(longStr))
+	}
+}
+
+func BenchmarkWriteStringOptimized(b *testing.B) {
+	buf := NewStringBuffer()
+	defer buf.Release()
+	longStr := "This is a test string for benchmarking the write performance of optimized string writing"
+	
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		buf.buf.Reset()
+		WriteString(&buf.buf, longStr)
+	}
+}
+
+func BenchmarkWriteStringDirect_Discard(b *testing.B) {
+	longStr := "This is a test string for benchmarking the write performance of direct byte slice conversion"
+	
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		ioutil.Discard.Write([]byte(longStr))
+	}
+}
+
+func BenchmarkWriteStringOptimized_Discard(b *testing.B) {
+	longStr := "This is a test string for benchmarking the write performance of optimized string writing"
+	
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		WriteString(ioutil.Discard, longStr)
+	}
+}
diff --git a/utility.go b/utility.go
new file mode 100644
index 0000000..b32d26b
--- /dev/null
+++ b/utility.go
@@ -0,0 +1,42 @@
+package twig
+
+import (
+	"bytes"
+	"io"
+	"sync"
+)
+
+// byteBufferPool is used to reuse byte buffers during node rendering
+var byteBufferPool = sync.Pool{
+	New: func() interface{} {
+		return &bytes.Buffer{}
+	},
+}
+
+// GetByteBuffer gets a bytes.Buffer from the pool
+func GetByteBuffer() *bytes.Buffer {
+	buf := byteBufferPool.Get().(*bytes.Buffer)
+	buf.Reset() // Clear any previous content
+	return buf
+}
+
+// PutByteBuffer returns a bytes.Buffer to the pool
+func PutByteBuffer(buf *bytes.Buffer) {
+	byteBufferPool.Put(buf)
+}
+
+// WriteString optimally writes a string to a writer
+// This avoids allocating a new byte slice for each string written
+func WriteString(w io.Writer, s string) (int, error) {
+	// Fast path for strings.Builder, bytes.Buffer and similar structs that have WriteString
+	if sw, ok := w.(io.StringWriter); ok {
+		return sw.WriteString(s)
+	}
+	
+	// Fallback path - reuse buffer from pool to avoid allocation
+	buf := GetByteBuffer()
+	buf.WriteString(s)
+	n, err := w.Write(buf.Bytes())
+	PutByteBuffer(buf)
+	return n, err
+}
\ No newline at end of file