Spring is here, and so are some new advanced models for our "I'm All In" tier! You now have a total of six different models to choose from. Noromaid 45B: This model is a fine-tuned version of Mixtral that excels at Role Playing and ERP. It features 8192 tokens of memory. You can experiment with a higher temperature on this model, as the default of 0.7 might be too low (recommended 1.0). Midnight Rose 70B: This model is a merge based of Airoboros designed for roleplaying and storytelling. It aims to produce lengthy outputs by default and is very creative. This model is limited to 4096 tokens and is much slower than our other models.

These newer models are only available on the web and not in our mobile apps. Your subscription can be used on all platforms, regardless of how you paid.

Also, when adding new models, it's challenging for us to plan the right GPU capacity to run these models. If one of these newer models becomes too slow, we suggest switching back to the default and trying again at a later time.

Happy Experimenting!