Update README.md

README.md

@@ -27,6 +27,8 @@ The training dataset was a private collection of around 2000 jokes. This model h
 
 See the example in the inference widget or How to use section for what constitues a narrative-style joke.
 
+For a slightly more accurate model (0.4% more) that is 65% slower at inference, see the [Deberta-v3 model](https://huggingface.co/Reggie/DeBERTa-v3-base-joke_detector). For a much more inaccurate model (2.4% less) that is way faster at inference, see the [distilbert model](https://huggingface.co/Reggie/distilbert-joke_detector).
+
 ### Install these first
 You'll need to pip install transformers & maybe sentencepiece