We've just made available the large language model Qwen3 235B which was recently released by Alibaba. This model supports 119 languages and is now available for our I'm All In tier with a context memory of 16K. Based on our Beta Testers, some of the model strengths are: 

• Rich Language & Realism: Many users praised its human-like dialogue, emotional articulation, and vivid descriptive language. It excels with internal monologues and detailed NSFW content.
• Character Fidelity: It handles bot personalities well, including accents, quirks, and backstories. Some users noted it was the first model to fully reflect deep character traits and physical details (like tattoos).
• Multi-character Capability: Several users found that Qwen3 could manage multiple characters in a scene effectively—responding as different personas when prompted properly or coded in.
• Creativity: The model was noted for surprising analogies, dramatic flair, and chaotic or comedic character portrayals (e.g., Deadpool types).
• NSFW Performance: Repeatedly praised for being slow, careful, and immersive in NSFW scenes—some preferred it over XL and DeepSeek for this purpose.
• Memory Use: Strong long-context memory made it better at keeping track of ongoing stories than some alternatives.

 Enjoy!