diff --git a/Sources/Tokenizers/Tokenizer.swift b/Sources/Tokenizers/Tokenizer.swift
index e1fe2530..fef67abe 100644
--- a/Sources/Tokenizers/Tokenizer.swift
+++ b/Sources/Tokenizers/Tokenizer.swift
@@ -116,6 +116,7 @@ struct TokenizerModel {
         "PreTrainedTokenizer": BPETokenizer.self,
         "Qwen2Tokenizer": BPETokenizer.self,
         "WhisperTokenizer": BPETokenizer.self,
+        "XLMRobertaTokenizer": UnigramTokenizer.self,
     ]
 
     static func unknownToken(from tokenizerConfig: Config) -> String? {
diff --git a/Tests/TokenizersTests/TokenizerTests.swift b/Tests/TokenizersTests/TokenizerTests.swift
index 821df186..72db5a4b 100644
--- a/Tests/TokenizersTests/TokenizerTests.swift
+++ b/Tests/TokenizersTests/TokenizerTests.swift
@@ -120,6 +120,20 @@ class PhiSimpleTests: XCTestCase {
     }
 }
 
+class RobertaTokenizerTests: XCTestCase {
+    /// https://github.com/huggingface/swift-transformers/issues/99
+    func testRobertaXLMTokenizer() async throws {
+        guard let tokenizer = try await AutoTokenizer.from(pretrained: "intfloat/multilingual-e5-small") as? PreTrainedTokenizer else {
+            XCTFail()
+            return
+        }
+
+        let ids = tokenizer.encode(text: "query: how much protein should a female eat")
+        let expected = [0, 41, 1294, 12, 3642, 5045, 21308, 5608, 10, 117776, 73203, 2]
+        XCTAssertEqual(ids, expected)
+    }
+}
+
 class UnregisteredTokenizerTests: XCTestCase {
     func testNllbTokenizer() async throws {
         do {