KaraKaraWarehouse/Llama-MiraiFanfare-3.3-70B
Llama-MiraiFanfare-3.3-70B by KaraKaraWitch is a 70 billion parameter language model, merged using the TIES method with EVA-LLaMA-3.33-70B-v0.0 as its base. This model is specifically tweaked to reduce verbosity and enhance creative output, offering a 32K context length. It is optimized for creative text generation, providing a more steerable alternative to models like Qwen2.5 for imaginative tasks.
Loading preview...
Overview
Llama-MiraiFanfare-3.3-70B is a 70 billion parameter language model developed by KaraKaraWitch, built upon the EVA-LLaMA-3.33-70B-v0.0 base model. This iteration, merged using the TIES method, focuses on refining the output to be less verbose and more creative, addressing a common frustration with overly wordy models. It incorporates components from Blackroot/Mirai-70B-1.0 and EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0 with specific density and weight parameters.
Key Capabilities
- Enhanced Creativity: Designed to produce more imaginative and less verbose text compared to its predecessors.
- Reduced Verbosity: Tweaked to avoid overly wordy responses, improving conciseness.
- Merge Architecture: Utilizes the TIES merge method, combining strengths of multiple models.
- ChatML Prompting: Recommended to be used with ChatML for optimal performance, contrary to typical Llama 3 model prompting.
Good For
- Creative Text Generation: Ideal for applications requiring imaginative and original content.
- Roleplay and Storytelling: Its focus on creativity makes it suitable for generating engaging narratives.
- Users Seeking Less Verbose Output: A good choice for those who find other models too verbose and hard to steer.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.