# Using pipelines versus InferenceClient

**URL:** <https://discuss.huggingface.co/t/using-pipelines-versus-inferenceclient/59460>\
**Category:** Beginners\
**Created:** [October 22, 2023, 5:23pm UTC](https://discuss.huggingface.co/t/using-pipelines-versus-inferenceclient/59460 "2023-10-22T17:23:12Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![mish2023](https://avatars.discourse-cdn.com/v4/letter/m/34f0e0/32.png) [@mish2023](https://discuss.huggingface.co/u/mish2023)\
**Post date:** [October 22, 2023, 5:23pm UTC](https://discuss.huggingface.co/t/using-pipelines-versus-inferenceclient/59460/1 "2023-10-22T17:23:12Z")

</div>

Hello, I see I can use a conversational model either via pipelines or via the InferenceClient. I would like to know best practice and if there is a recommendation on what to use from these two. The third way (via transformers - model and tokenizer) is harder and I dont want to use that. Thank you!

### Method 1 : Pipelines

### This also indicates the model it defaults to

```auto
from transformers import pipeline, Conversation
converse = pipeline("conversational")

conversation_1 = Conversation("Going to the movies tonight - any suggestions?")
converse([conversation_1])

```

### Method 2 : InferenceClient

```auto
from huggingface_hub import InferenceClient
client = InferenceClient()
output = client.conversational("Going to the movies tonight - any suggestions?")
response = output['generated_text']
print(response)

# This will work once a PR is merged
# https://github.com/huggingface/huggingface_hub/pull/1770
default_model = client.get_recommended_model()
print(default_model)

```