We’re excited to share the release of our second-generation model: Nexus-GPT2-V2. It’s built to deliver more natural, engaging conversations—especially for role-play, companionship, and creative storytelling.

Unlike traditional models that prioritize raw reasoning, Nexus-GPT2-V2 focuses on personality, tone, and emotional nuance. The result is a model that feels more human-like, expressive, and alive.

This version is a big step up from the original:

  • Expanded from 124M to 355M parameters
  • Training data increased from 400k to 500k messages, with better filtering and clearer context
  • Stronger ability to understand intent, sustain longer interactions, and maintain a consistent personality

We’ve also fine-tuned how the model adapts to tone. Whether you’re casually chatting, diving into immersive role-play, or exploring expressive dialogue, Nexus-GPT2-V2 can shift its voice to fit the moment.

And we’re not stopping here. Our team is already working on the next generation—Nexus-GPT3-V3—trained on even larger datasets with better filtering. But for now, we invite you to try out Nexus-GPT2-V2 and experience the improvements firsthand.