diff --git a/bindings/python/src/tokenizer.rs b/bindings/python/src/tokenizer.rs
index 75c1a175..3b9b6cac 100644
--- a/bindings/python/src/tokenizer.rs
+++ b/bindings/python/src/tokenizer.rs
@@ -233,6 +233,10 @@ impl Tokenizer {
         Ok(self.tokenizer.add_tokens(&tokens))
     }
 
+    fn add_special_tokens(&mut self, tokens: Vec<&str>) -> PyResult<usize> {
+        Ok(self.tokenizer.add_special_tokens(&tokens))
+    }
+
     fn train(&mut self, trainer: &Trainer, files: Vec<String>) -> PyResult<()> {
         trainer.trainer.execute(|trainer| {
             if let Err(e) = self.tokenizer.train(trainer, files) {
diff --git a/tokenizers/src/tokenizer/mod.rs b/tokenizers/src/tokenizer/mod.rs
index 631d6293..9603eebd 100644
--- a/tokenizers/src/tokenizer/mod.rs
+++ b/tokenizers/src/tokenizer/mod.rs
@@ -140,6 +140,7 @@ pub struct Tokenizer {
     added_tokens: HashMap<AddedToken, u32>,
     added_tokens_r: HashMap<u32, AddedToken>,
     split_re: Option<regex::Regex>,
+    special_tokens: HashMap<String, u32>,
 
     // General processing parameters
     trunc: Option<TruncationParams>,
@@ -159,6 +160,7 @@ impl Tokenizer {
             added_tokens: HashMap::new(),
             added_tokens_r: HashMap::new(),
             split_re: None,
+            special_tokens: HashMap::new(),
 
             trunc: None,
             padding: None,
@@ -483,6 +485,24 @@ impl Tokenizer {
         Ok(final_encoding)
     }
 
+    /// Register the given tokens as special tokens. This is especially useful for removing
+    /// these special tokens while decoding
+    pub fn add_special_tokens(&mut self, tokens: &[&str]) -> usize {
+        let added_tokens = tokens
+            .iter()
+            .map(|t| AddedToken::from((*t).to_owned()))
+            .collect::<Vec<_>>();
+
+        let added = self.add_tokens(&added_tokens);
+        for token in tokens {
+            if let Some(id) = self.token_to_id(token) {
+                self.special_tokens.entry((*token).to_owned()).or_insert(id);
+            }
+        }
+
+        added
+    }
+
     /// Add the given tokens to the added vocabulary
     pub fn add_tokens(&mut self, tokens: &[AddedToken]) -> usize {
         let mut ignored = 0;