L3.3-Mokume-Gane-R1-70b-v1.1

Created by SteelSkull →

Support on Ko-fi

Model Information

L3.3-Mokume-Gane-R1-70b-v1.1 | v0.5.C

L3.3 = Llama 3.3 SCE Merge R1 = Deepseek R1 70b Parameters v0.5.C

Model Composition

• L3.1x3.3-DS-Hydroblated-R1-70B-v4.1 Base model
• EVA-LLaMA-3.33-v0.0 Core capabilities
• Euryale-v2.3 Enhanced reasoning
• Cirrus-x1 Improved coherence
• Hanami-x1 Balanced responses
• Anubis-v1 Enhanced detail
• Negative_LLAMA Reduced bias

Model Series

Mokume-Gane is part of a three-model experimental series, each offering distinct characteristics:

• L3.3-San-Mai-R1-70b - The original model (OG), Gold Tier
• L3.3-Cu-Mai-R1-70b - Version A, OG with a twist
• L3.3-Mokume-Gane-R1-70b-v1.1 - Version C, emphasizing creativity

Model Info

Named after the Japanese metalworking technique 'Mokume-gane' (木目金), meaning 'wood grain metal', this model embodies the artistry of creating distinctive layered patterns through the careful mixing of different components. Just as Mokume-gane craftsmen blend various metals to create unique visual patterns, this model combines specialized AI components to generate creative and unexpected outputs.

Technical Architecture

Built on the DS-Hydroblated-R1 foundation and utilizing the SCE (Select, Calculate, and Erase) merge method, Mokume-Gane-R1 integrates components from multiple high-performance models:

• EVA and EURYALE foundations provide the creative expression and scene comprehension
• Cirrus and Hanami elements enhance reasoning while maintaining balance
• Anubis components deliver detailed scene descriptions
• Negative_LLAMA integration helps maintain perspective and reduce bias

Core Philosophy

Mokume-Gane-R1 embraces the spirit of creative exploration while maintaining technical precision. Community feedback consistently highlights its ability to generate unique and unexpected outputs, setting it apart as the "most creative" model in its class. While it excels in character adherence and creative expression, it requires careful tuning to balance its innovative tendencies with coherent output.

Prompting and Reasoning

A key feature of this model series is its enhanced reasoning capabilities, which can be effectively triggered through structured prompting. By incorporating clear logical frameworks and step-by-step thinking patterns in prompts, users can unlock deeper analytical responses and more coherent reasoning chains. This is particularly effective when combined with the model's creative tendencies, allowing for both innovative and well-reasoned outputs.

Community Insights

The model has received notable praise for its creativity and unique outputs, with users particularly highlighting its:

• Exceptional creativity and ability to generate novel responses
• Strong character adherence and natural dialogue flow
• Distinctive outputs that differentiate it from standard models
• Potential for outstanding results with proper sampler tuning

However, users should note that the model's creative strength comes with certain considerations:

• Results can be variable, ranging from exceptional to requiring refinement
• May require careful prompt engineering to maintain focus
• Benefits significantly from appropriate sampler settings

Base Architecture

At its core, Mokume-Gane-R1 utilizes the custom DS-Hydroblated-R1 base model, engineered for stability and enhanced reasoning. The SCE merge method, with settings finely tuned based on extensive community feedback, enables precise component integration while maintaining model coherence and reliability. This creates a foundation that supports both creative expression and technical accuracy.

Recommended Sampler Settings: By @Geechan

Static Temperature:

1 - 1.05

Min P

0.03

DRY Settings: (optional)

Multiplier 0.8

Base 1.75

Length 4

Recommended Templates & Prompts

LLam@ception → by @.konnect

LeCeption → by @Steel > A completly revamped XML version of Llam@ception 1.5.2 with stepped thinking and Reasoning added

LECEPTION REASONING CONFIGURATION:

Start Reply With:

'<think> OK, as an objective, detached narrative analyst, let's think this through carefully:'

Reasoning Formatting (no spaces):

Prefix: '<think>'

Suffix: '</think>'

Support & Community:

Support on Ko-fi Join Discord

....Special Thanks

• @Geechan for feedback and sampler settings
• @Konnect for their feedback and templates
• @Kistara for their feedback and help with the model mascot design
• @Thana Alt for their feedback and Quants
• @Lightning_missile for their feedback
• The Arli community for feedback and testers
• The BeaverAI communty for feedback and testers

I wish I could add everyone but im pretty sure it would be as long as the card!