Llama-3.3-70B-Cu-Mai-R1

Creative Model

View on Hugging FaceBack to Models

Hourly Usage

Performance Metrics

Avg. Total Time

28.20s

Avg. TTFT

13.85s

Avg. Prefill TPS

2492.49

Avg. Gen TPS

17.89

Model Information

Context Size

32768

Quantization

r64

Engine

aphrodite

Creation Method

Merge

Model Type

Llama70B

Chat Template

Llama 3

Reasoning

No

Vision

No

Parameters

70B

Added At

2/21/2025


license: llama3.3 base_model:

  • TheSkullery/L3.1x3.3-DS-Hydroblated-R1-70B-v4.1 pipeline_tag: text-generation

L3.3-Cu-Mai-R1-70b

Model banner

-Model Information

L3.3-Cu-Mai-R1-70b

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

Model Composition

Model Info

Cu-Mai, a play on San-Mai for Copper-Steel Damascus, represents a significant evolution in the three-part model series alongside San-Mai (OG) and Mokume-Gane. While maintaining the grounded and reliable nature of San-Mai, Cu-Mai introduces its own distinct "flavor" in terms of prose and overall vibe. The model demonstrates strong adherence to prompts while offering a unique creative expression.

Technical Architecture

L3.3-Cu-Mai-R1-70b integrates specialized components through the SCE merge method:

  • EVA and EURYALE foundations for creative expression and scene comprehension
  • Cirrus and Hanami elements for enhanced reasoning capabilities
  • Anubis components for detailed scene description
  • Negative_LLAMA integration for balanced perspective and response

User Experience & Capabilities

Users consistently praise Cu-Mai for its:

  • Exceptional prose quality and natural dialogue flow
  • Strong adherence to prompts and creative expression
  • Improved coherency and reduced repetition
  • Performance on par with the original model

While some users note slightly reduced intelligence compared to the original, this trade-off is generally viewed as minimal and doesn't significantly impact the overall experience. The model's reasoning capabilities can be effectively activated through proper prompting techniques.

Model Series Context

Cu-Mai (Version A) is part of a three-model series:

  • L3.3-San-Mai-R1-70b (OG model) - The original foundation
  • L3.3-Cu-Mai-R1-70b (Version A) - Enhanced creative expression
  • L3.3-Mokume-Gane-R1-70b (Version C) - Distinct variation with unique characteristics

Base Architecture

At its core, L3.3-Cu-Mai-R1-70b utilizes the entirely custom Hydroblated-R1 base model, specifically engineered for stability, enhanced reasoning, and performance. The SCE merge method, with settings finely tuned based on community feedback from evaluations of Experiment-Model-Ver-0.5, Experiment-Model-Ver-0.5.A, Experiment-Model-Ver-0.5.B, Experiment-Model-Ver-0.5.C, Experiment-Model-Ver-0.5.D, L3.3-Exp-Nevoria-R1-70b-v0.1 and L3.3-Exp-Nevoria-70b-v0.1, enables precise and effective component integration while maintaining model coherence and reliability.

Recommended Sampler Settings: By @Geechan

Static Temperature:

1 - 1.05

Min P

0.02

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:

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!