diff --git a/pdm.lock b/pdm.lock
index 8278b7b..467e81b 100644
--- a/pdm.lock
+++ b/pdm.lock
@@ -5,7 +5,7 @@
 groups = ["default", "dev"]
 strategy = ["cross_platform", "inherit_metadata"]
 lock_version = "4.4.1"
-content_hash = "sha256:ba436b1fde50163f18ee08183fcedc49fa36b2c6dc2d4ac1a005056c052128dd"
+content_hash = "sha256:c9d1751b45e83e3705d56d669dcaa4a911db6fdf591aa53f15f7bb5aec672e5d"
 
 [[package]]
 name = "appdirs"
diff --git a/src/word_prediction/__init__.py b/src/word_prediction/__init__.py
index 9a1b7b2..3af85f9 100644
--- a/src/word_prediction/__init__.py
+++ b/src/word_prediction/__init__.py
@@ -1,3 +1,5 @@
+from dataclasses import dataclass
+from typing import Optional, Tuple
 from sparv.api import (  # type: ignore [import-untyped]
     annotator,
     Output,
@@ -10,26 +12,16 @@
 from transformers import (  # type: ignore [import-untyped]
     BertTokenizer,
     BertForMaskedLM,
-    FillMaskPipeline,
 )
+from word_prediction.predictor import HuggingFaceTopKPredictor
 
-__description__ = "Calculating word neighbours by mask a word in a BERT model."
+__description__ = "Calculating word predictions by mask a word in a BERT model."
 
 
 __config__ = [
     Config(
-        "word_prediction.model",
-        description="Huggingface pretrained model name",
-        default="KBLab/bert-base-swedish-cased",
-    ),
-    Config(
-        "word_prediction.tokenizer",
-        description="HuggingFace pretrained tokenizer name",
-        default="KBLab/bert-base-swedish-cased",
-    ),
-    Config(
-        "word_prediction.num_neighbours",
-        description="The number of neighbours to list",
+        "word_prediction.num_predictions",
+        description="The number of predictions to list",
         default=5,
     ),
 ]
@@ -41,36 +33,83 @@
 TOK_SEP = " "
 
 
+@dataclass
+class HuggingfaceModel:
+    model_name: str
+    model_revision: str
+    tokenizer_name: Optional[str] = None
+    tokenizer_revision: Optional[str] = None
+
+    def tokenizer_name_and_revision(self) -> Tuple[str, str]:
+        if tokenizer_name := self.tokenizer_name:
+            return tokenizer_name, self.tokenizer_revision or "main"
+        else:
+            return self.model_name, self.model_revision
+
+
+MODELS = {
+    "kb-bert": HuggingfaceModel(
+        model_name="KBLab/bert-base-swedish-cased",
+        model_revision="c710fb8dff81abb11d704cd46a8a1e010b2b022c",
+    )
+}
+
+
 @annotator(
-    "Word neighbour tagging with a masked Bert model",
+    "Word prediction tagging with a masked Bert model",
 )
-def annotate_masked_bert(
-    out_neighbour: Output = Output(
-        "<token>:word_prediction.transformer-neighbour",
-        cls="transformer_neighbour",
-        description="Transformer neighbours from masked BERT (format: '|<word>:<score>|...|)",
+def predict_words__kb_bert(
+    out_prediction: Output = Output(
+        "<token>:word_prediction.word-prediction--kb-bert",
+        cls="word_prediction",
+        description="Word predictions from masked BERT (format: '|<word>:<score>|...|)",
     ),
     word: Annotation = Annotation("<token:word>"),
     sentence: Annotation = Annotation("<sentence>"),
-    model_name: str = Config("word_prediction.model"),
-    tokenizer_name: str = Config("word_prediction.tokenizer"),
-    num_neighbours_str: str = Config("word_prediction.num_neighbours"),
+    num_predictions_str: str = Config("word_prediction.num_predictions"),
 ) -> None:
-    logger.info("annotate_masked_bert")
+    logger.info("predict_words")
     try:
-        num_neighbours = int(num_neighbours_str)
+        num_predictions = int(num_predictions_str)
     except ValueError as exc:
         raise SparvErrorMessage(
-            f"'word_prediction.num_neighbours' must contain an 'int' got: '{num_neighbours_str}'"
+            f"'word_prediction.num_predictions' must contain an 'int' got: '{num_predictions_str}'"
         ) from exc
-    tokenizer = BertTokenizer.from_pretrained(tokenizer_name)
-    model = BertForMaskedLM.from_pretrained(model_name)
+    tokenizer_name, tokenizer_revision = MODELS["kb-bert"].tokenizer_name_and_revision()
+
+    tokenizer = BertTokenizer.from_pretrained(
+        tokenizer_name, revision=tokenizer_revision
+    )
+    model = BertForMaskedLM.from_pretrained(
+        MODELS["kb-bert"].model_name, revision=MODELS["kb-bert"].model_revision
+    )
 
-    hf_top_k_predictor = HuggingFaceTopKPredictor(model=model, tokenizer=tokenizer)
+    predictor = HuggingFaceTopKPredictor(model=model, tokenizer=tokenizer)
 
     sentences, _orphans = sentence.get_children(word)
     token_word = list(word.read())
-    out_neighbour_annotation = word.create_empty_attribute()
+    out_prediction_annotation = word.create_empty_attribute()
+
+    run_word_prediction(
+        predictor=predictor,
+        num_predictions=num_predictions,
+        sentences=sentences,
+        token_word=token_word,
+        out_prediction_annotations=out_prediction_annotation,
+    )
+
+    logger.info("writing annotations")
+    out_prediction.write(out_prediction_annotation)
+
+
+def run_word_prediction(
+    predictor: HuggingFaceTopKPredictor,
+    num_predictions: int,
+    sentences,
+    token_word: list,
+    out_prediction_annotations,
+) -> None:
+    logger.info("run_word_prediction")
 
     logger.progress(total=len(sentences))  # type: ignore
     for sent in sentences:
@@ -86,27 +125,7 @@ def annotate_masked_bert(
                 for token_index in sent
             )
 
-            neighbours_scores = hf_top_k_predictor.get_top_k_predictions(
-                sent_to_tag, k=num_neighbours
-            )
-            out_neighbour_annotation[token_index_to_mask] = neighbours_scores
-
-    logger.info("writing annotations")
-    out_neighbour.write(out_neighbour_annotation)
-
-
-class HuggingFaceTopKPredictor:
-    def __init__(self, *, tokenizer, model) -> None:
-        self.tokenizer = tokenizer
-        self.model = model
-        self.pipeline = FillMaskPipeline(model=model, tokenizer=tokenizer)
-
-    def get_top_k_predictions(self, text: str, k=5) -> str:
-        if predictions := self.pipeline(text, top_k=k):
-            predictions_str = "|".join(
-                f"{pred['token_str']}:{pred['score']}" # type: ignore
-                for pred in predictions 
+            predictions_scores = predictor.get_top_k_predictions(
+                sent_to_tag, k=num_predictions
             )
-            return f"|{predictions_str}|"
-        else:
-            return "|"
+            out_prediction_annotations[token_index_to_mask] = predictions_scores
diff --git a/src/word_prediction/predictor.py b/src/word_prediction/predictor.py
new file mode 100644
index 0000000..b046d41
--- /dev/null
+++ b/src/word_prediction/predictor.py
@@ -0,0 +1,20 @@
+from transformers import (  # type: ignore [import-untyped]
+    FillMaskPipeline,
+)
+
+
+class HuggingFaceTopKPredictor:
+    def __init__(self, *, tokenizer, model) -> None:
+        self.tokenizer = tokenizer
+        self.model = model
+        self.pipeline = FillMaskPipeline(model=model, tokenizer=tokenizer)
+
+    def get_top_k_predictions(self, text: str, k=5) -> str:
+        if predictions := self.pipeline(text, top_k=k):
+            predictions_str = "|".join(
+                f"{pred['token_str']}:{pred['score']}"  # type: ignore
+                for pred in predictions
+            )
+            return f"|{predictions_str}|"
+        else:
+            return "|"