diff --git a/Cargo.toml b/Cargo.toml
index 30e73eec..d3cbf3ab 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "base64"
-version = "0.13.0"
+version = "0.13.1"
 authors = ["Alice Maz <alice@alicemaz.com>", "Marshall Pierce <marshall@mpierce.org>"]
 description = "encodes and decodes base64 as bytes or utf8"
 repository = "https://github.com/marshallpierce/rust-base64"
diff --git a/RELEASE-NOTES.md b/RELEASE-NOTES.md
index 1048c1e8..6a2e3cae 100644
--- a/RELEASE-NOTES.md
+++ b/RELEASE-NOTES.md
@@ -1,3 +1,7 @@
+# 0.13.1
+
+- More precise decode buffer sizing, avoiding unnecessary allocation in `decode_config`.
+
 # 0.13.0
 
 - Config methods are const
diff --git a/src/decode.rs b/src/decode.rs
index 4cc937d5..e349240a 100644
--- a/src/decode.rs
+++ b/src/decode.rs
@@ -107,7 +107,14 @@ pub fn decode<T: AsRef<[u8]>>(input: T) -> Result<Vec<u8>, DecodeError> {
 ///```
 #[cfg(any(feature = "alloc", feature = "std", test))]
 pub fn decode_config<T: AsRef<[u8]>>(input: T, config: Config) -> Result<Vec<u8>, DecodeError> {
-    let mut buffer = Vec::<u8>::with_capacity(input.as_ref().len() * 4 / 3);
+    let decoded_length_estimate = (input
+        .as_ref()
+        .len()
+        .checked_add(3)
+        .expect("decoded length calculation overflow"))
+        / 4
+        * 3;
+    let mut buffer = Vec::<u8>::with_capacity(decoded_length_estimate);
 
     decode_config_buf(input, config, &mut buffer).map(|_| buffer)
 }
@@ -870,4 +877,17 @@ mod tests {
             }
         }
     }
+
+    #[test]
+    fn decode_config_estimation_works_for_various_lengths() {
+        for num_prefix_quads in 0..100 {
+            for suffix in &["AA", "AAA", "AAAA"] {
+                let mut prefix = "AAAA".repeat(num_prefix_quads);
+                prefix.push_str(suffix);
+                // make sure no overflow (and thus a panic) occurs
+                let res = decode_config(prefix, STANDARD);
+                assert!(res.is_ok());
+            }
+        }
+    }
 }
diff --git a/src/write/encoder.rs b/src/write/encoder.rs
index 8a48f438..4bb57eb0 100644
--- a/src/write/encoder.rs
+++ b/src/write/encoder.rs
@@ -223,7 +223,7 @@ impl<W: Write> Write for EncoderWriter<W> {
     /// Under non-error circumstances, this returns `Ok` with the value being the number of bytes
     /// of `input` consumed. The value may be `0`, which interacts poorly with `write_all`, which
     /// interprets `Ok(0)` as an error, despite it being allowed by the contract of `write`. See
-    /// https://github.com/rust-lang/rust/issues/56889 for more on that.
+    /// <https://github.com/rust-lang/rust/issues/56889> for more on that.
     ///
     /// If the previous call to `write` provided more (encoded) data than the delegate writer could
     /// accept in a single call to its `write`, the remaining data is buffered. As long as buffered