Models

microsoft/WizardLM 2 8x22B

microsoft/WizardLM-2-8x22B: Bewitched

WizardLM 2 8x22b stands out as the best model for applications. It delivers precise and comprehensive answers to knowledge-based questions and excels in inferential reasoning and mathematical problem-solving, outperforming all other models I’ve tested.This state-of-the-art large language model, developed by Microsoft AI, showcases enhanced capabilities in complex chat, multilingual tasks, reasoning, and agent-based use cases. It is also the most advanced open-source LLM, ranking highest in Microsoft’s internal evaluations for highly complex tasks.

 

rAIfle/SorcererLM 8x22b bf16: Bewildered

Sorcerer model is a fine-tuned version of WizardLM 2 8x22B, optimized specifically for role-playing (RP) applications. By applying a low-rank adaptation (LoRA) with parameters r=16 and alpha=32, and training over two epochs on a curated dataset of c2-logs, Sorcerer enhances the original model’s performance in generating contextually rich and stylistically appropriate responses for RP scenarios.

 

Key Differentiation

Unlike WizardLM 2 8x22B, which serves general applications, Sorcerer is tailored specifically for role-playing contexts. This specialization makes it particularly adept at generating immersive and engaging narratives, character dialogues, and scenarios that align with the dynamic nature of role-playing environments.

 

Want to Try It?

Experience WizardLM 2 8x22b and SorcererLM 8x22b on the following platforms, both offering a 32K context window:

 

 

Settings Bank

All the settings you need in one place!

 

Creative

Catbox link: https://files.catbox.moe/j8895z.json

Tokenizer: Mistral

(Yes you can use the same settings for both models)
Parameter Value
 temp 1.2
 temperature_last true
 top_p 1
 top_k -1
 top_a 0
 typical_p 0.7
 min_p 0
 rep_pen 1
 min_length 0
 freq_pen 0
 presence_pen 0.5
Created by GERGE

 

Wild and Stable
Catbox link: https://files.catbox.moe/hq2xct.json
Parameter Value
 temp 1
 temperature_last true
 top_p 1
 top_k -1
 top_a 0
 typical_p 1
 min_p 0.02
 rep_pen 1
 min_length 0
 freq_pen 0
 presence_pen 0.5
Created by GERGE

 

Looking for Similar Models?

Explore alternatives:

Review of Sao10K/L3 70B Euryale v2.1
Click here

Review of Infermatic/MN 12B Inferor v0.0
Click here

Want to Know More?

Have questions or want to explore settings, examples, or community experiences? Join the discussion on our Discord server! Click the button below to connect: