Ultra GPT 4.1
Ultra GPT 4.1 is one of the Ultra Models offered by CrushOn for its premium subscribers, offering superior reply speed and performance at the cost of higher coin price per message. Provided by OpenAI, it is derived from the LLM ChatGPT is based on, giving the model an incredibly vast amount of factual knowledge that it could use to its advantage in immersive roleplaying. The 4.1 models offers improved coherence in long conversations and better logical flow, less predictable and stiff than older GPT models.
- It also comes with a non-VIP version in the form of GPT 4.1 Mini, a lightweight version of Ultra GPT 4.1, engineered for fast performance with minimal resource usage.
| Free 8K | Pro 8K | Pro 16K | Pro 24K |
|---|---|---|---|
| ✓ (Mini Only) | ✓ | ✓ | ✓ (Ultra Only) |
From here on out, the content may not be fully accurate to each user and act only as general reference. Each user often could see highly different experiences as there are many variables that affect the behavior and quality of the models. From profile cards, the style in which the user converse with the character, the set parameters for the model, and of course the personality of the characters. And periodically, even minimal adjustments made to the model by the developers could also substantially change its behavior whether intended or not.
Strengths and Flaws
Many of the GPT models are often notorious for being 'obviously ChatGPT'. However, the GPT 4.1 models have been repurposed to alleviate that issue. Its strength comes from free flowing and creative responses that feels much more natural and adheres well to character traits and personalities, not just the character's but the user's own traits that are set on the profile card. It is also quite flexible in the way that it would match the tone and style of the user's roleplay method, often mimicking formatting and becoming naturally more uniform with the user's own chat over time.
GPT's greatest flaw, now remain with its highly strict filtering and, at times, formatting inconsistencies, especially with its use of asterisks. These are both problems shared by the Ultra and Mini version of the model.
- Unique to the Mini version however, there would periodically be logic gaps in its responses that may range from subtle to jarring. From misunderstanding context of user's speech to misunderstanding the setting of time and location.
Common Issues with the Model
Due to the strength of GPT's filtering, it may be impossible at times to try and force the model into NSFW scenarios, even with custom prompting and instructions.
- To this end, switching to a reliable jailbreak model such as CrushOn Leo is often a great way to train Leo for the interaction to remain close to the style GPT 4.1 produce and once any NSFW scene is over, one could safely return back to the GPT 4.1 model.
From time to time, GPT 4.1 would also have its own preferrable method of roleplay, using asterisks for impersonal narration and not using quotations for their dialogue. Understandably, this way of formatting might not be for everyone.
- It is important to note, however, that the model is also referencing the user's own input for its responses. At times, being consistent in your own replies would be enough to teach the model your preferred method of roleplay and formatting.
- If that doesn't work, manually editing the model's response should also easily teach the model to use that edited version of the format quite consistently.
Example of Use
On the right, you could see the example of the models (Ultra and Mini) being used regularly with the OC Yumi by Since.