Skip to main content
Browse Models

TheDrummer

Anubis 70B v1.1

Released

2025-06-17

Family

Llama 3

Type

Fine-Tuned Model

Huggingface Page

Downloads

External Download

You are about to open a link to an external source. Verify the URL before continuing.

Download

4-bit GGUF (Q4_K_M)

GGUF · TheDrummer_Anubis-70B-v1.1-Q4_K_M.gguf

Model Report

Overview

Anubis 70B v1.1, also known as "Shimmer Edition," is a large language model designed for generative AI tasks, with a particular emphasis on character consistency and dynamic dialogue. Developed by TheDrummer, Anubis 70B v1.1 builds upon its predecessor, Anubis 70B v1.0, with improvements in its ability to adhere to specific character traits and conversational nuances. The model uses fine-tuning techniques applied to Meta's Llama 3.3-70B-Instruct architecture, resulting in capabilities for creative text generation and interactive storytelling.

A stylized digital illustration of a futuristic spaceship against a shimmering cosmic background, serving as the emblematic image for Anubis 70B v1.1 Shimmer Edition.

Figure 1. Anubis 70B v1.1 Shimmer Edition is represented by a futuristic spaceship in a cosmic environment, reflecting the model's unique branding and capabilities.

Full sizeSource

Development and Model Architecture

Anubis 70B v1.1 is a product of fine-tuning on top of the Llama 3.3-70B-Instruct model, which itself derives from the Llama 3.1-70B architecture. The model contains approximately 70.6 billion parameters and utilizes the BF16 tensor format, enabling efficient computation and inference at scale. This fine-tuning process was designed to enhance aspects such as character adherence and to allow for more nuanced text generation styles.

The "Shimmer Edition" designation signifies the distinct stylistic direction taken in this iteration. Unlike standard Llama derivatives, Anubis 70B v1.1 aims to offer a different approach to character-driven content, positioning itself as a tool for role-play and creative writing scenarios. According to the original source, users have observed that the model captures character mannerisms and dialogue shifts with higher fidelity than prior versions.

Training Approach and Data Foundations

While explicit details regarding the training datasets and methodologies used for Anubis 70B v1.1 are not available in the published information, it is confirmed that the model builds directly upon the Llama 3.3-70B-Instruct foundation. This fine-tuning process adapts general large language modeling capabilities to more targeted use cases, such as dialogue, persona adoption, and creativity in text output.

The model's training objectives centered on improving "unalignment," allowing for more open-ended and diverse responses beyond constrained outputs, as well as advancing its adherence to stylistic and behavioral patterns of defined characters. Empirical feedback from the community suggests that these adjustments facilitate more consistent characterization, even when interacting with elaborate or historically accurate character cards, broadening the scope for interactive fiction and narrative simulation.

Capabilities and Use Cases

A core strength of Anubis 70B v1.1 lies in its ability to simulate characters with consistent voice, behavior, and style across extended dialogues. This makes it suitable for applications where nuanced role-play and immersive storytelling are essential. The model demonstrates proficiency in maintaining distinct character voices, mannerisms, and contextual engagement—a trait valued for writers, interactive fiction platforms, and creative AI environments.

Common use cases highlighted for Anubis 70B v1.1 include:

  • Role-playing assistants capable of long-term persona retention.
  • Creative writing tools that generate dialogue and narrative with consistent characterization.
  • Interactive storytelling applications benefitting from dynamic, responsive characters.

Users have noted that Anubis 70B v1.1 brings a distinctive quality to its output, offering a contrasting writing style compared to previous Llama-based models and even to its own prior version.

Model Access, Instructions, and Deployment

Anubis 70B v1.1 is designed for compatibility with the Llama 3 Chat Template, enabling integration with conversational interfaces. Multiple format conversions are available for deployment across various inference environments, including GGUF, iMatrix, and EXL3 formats, as provided by the model's maintainers.

The model is publicly accessible for research and development through distribution platforms, facilitating both experimentation and academic study. Documentation for implementation, input formatting, and supported templates align with those prescribed for Llama 3-family models.

Model Family, Iterative Improvements, and Limitations

Anubis 70B v1.1 directly succeeds Anubis 70B v1.0, with user and developer feedback noting changes in its "character adherence" and output diversity. The update represents a modification of model behavior through targeted fine-tuning, yielding a different user experience.

As with many large generative models, explicit limitations are not outlined in the available sources. Performance may vary based on the complexity of requested character simulations, the length of contextual inputs, and application-specific requirements. License information for Anubis 70B v1.1 is not specified in the public documentation.

Helpful External Resources

For further exploration, technical materials, or community engagement, the following resources may be useful:

About Llama 3: The Llama 3 family of AI models, developed by Meta, represents a significant advancement in open-source large language models, offering parameter sizes up to 405 billion and supporting context windows of up to 128k tokens. Llama 3.1, 3.2, and 3.3 optimize this performance through distillation learning and improved multimodal capabilities.

More in the Llama 3 Family

Meta /

Llama 3 8B

Large language model with 8 billion parameters featuring transformer architecture, trained on 15 trillion tokens for text generation and coding tasks.
Meta /

Llama 3 70B

State-of-the-art 70B foundation model from Meta, trained on over 15 trillion tokens.
Meta /

Llama 3.1 8B

Llama 3.1 is a new state-of-the-art large language model from Meta.
Sao10K /

Llama 3.1 8B Stheno v3.4

An 8-billion parameter language model fine-tuned for multi-turn dialogue, creative writing, and roleplaying using curated conversational datasets and synthetic data.
Deepseek AI /

DeepSeek R1 Distill Llama 8B

Distilled 8B-parameter model optimized for mathematical reasoning and code generation through knowledge transfer from larger reinforcement learning-trained teacher models.
Deep Cogito /

Cogito V1 Preview 8B

A Llama 3.1-based model trained with Iterated Distillation and Amplification, featuring dual reasoning modes and tool calling capabilities.
Meta /

Llama 3.1 70B

The Llama 3.1 series of open models rivals top closed models in performance. It was trained on over 15 trillion tokens using over 16K H100 GPUs. These models display state-of-the-art capabilities in general knowledge, steerability, math, tool use, and translation.
Deep Cogito /

Cogito V1 Preview 70B

A 70B parameter instruction-tuned model based on Llama 3.1 architecture featuring dual reasoning modes and multilingual tool-calling capabilities.
Meta /

Llama 3.2 3B

The next iteration in the Llama series of open models. This lightweight model was designed to run on edge devices, even mobile.
Cognitive Computations /

Dolphin 3.0 Llama3.2 3B

An uncensored instruct-tuned 3.2B parameter language model that grants users full control over system prompts and behavioral alignment.
Deep Cogito /

Cogito V1 Preview 3B

A 3B-parameter multilingual instruction-tuned model based on Llama 3.2 that supports tool-calling and features dual operational modes for standard and extended reasoning.
Meta /

Llama 3.3 70B

Llama 3.3 is a text-only 70B instruction-tuned model that provides enhanced performance relative to Llama 3.1 70B and to Llama 3.2 90B when used for text-only applications. For some applications, Llama 3.3 70B approaches the performance of Llama 3.1 405B.
Sao10K /

L3.3 70B Euryale v2.3

A 70-billion parameter language model fine-tuned from Llama 3.3 for creative writing and role-playing applications using custom datasets.
Sao10K /

70B L3.3 Cirrus x1

A 70.6-billion parameter language model finetuned from Llama 3.3 using extended training and checkpoint merging techniques for improved output stability.
TheDrummer /

Anubis 70B v1

A 70.6-billion parameter text generation model fine-tuned from Llama 3.3, designed for creative writing and role-playing applications.
LatitudeGames /

Wayfarer Large 70B Llama 3.3

A 70.6-billion parameter language model fine-tuned for adventure role-play scenarios, emphasizing conflict, tension, and narrative stakes in second-person storytelling.
Deepseek AI /

DeepSeek R1 Distill Llama 70B

A 70B parameter dense language model distilled from DeepSeek-R1 using Llama 3.3 architecture, optimized for mathematical and coding reasoning tasks.

More from TheDrummer

TheDrummer /

Behemoth 123B v1.2

A 123-billion parameter language model optimized for conversational AI, creative prose generation, and role-playing applications with enhanced narrative consistency.
TheDrummer /

Cydonia 24B v2

A fine-tuned 23.6 billion parameter Mistral-based model designed for long-context conversations and maintaining narrative coherence across extended dialogues.
TheDrummer /

Rocinante 12B v1.1

A 12.2 billion parameter text generation model optimized for creative storytelling, role-playing scenarios, and adventure-based interactive fiction applications.

Compatible Apps

Open WebUI
open-webui /

Open WebUI

A polished, self-hosted chat interface for LLMs with Ollama integration, multimodal prompts, and extensive workspace customization.

Web UI

Chat UIs · Beginner Friendly

L
ggml-org /

llama.cpp

GGUF model inference with a polished web UI and OpenAI-format API. CPU-only build — high hardware compatibility, works on any machine without a GPU.

Web UI · API · CLI

LLM Inference · Chat UIs

L
ggml-org /

llama.cpp (CUDA)

GGUF model inference with a polished web UI and OpenAI-format API. CUDA build — GPU-accelerated for NVIDIA GPUs.

Web UI · API · CLI

LLM Inference · Chat UIs

Text Generation Web UI
oobabooga /

Text Generation Web UI

A feature-rich interface for running and experimenting with open-weight LLMs, including multiple inference backends, plugins, and tuning controls.

Web UI · API

Chat UIs · LLM Inference