LiamVisionary/swarm-sovereign-12b

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 4, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

LiamVisionary/swarm-sovereign-12b is a 12 billion parameter Gemma 4-based instruction-tuned chat model, prepared for local inference using GGUF and Hugging Face Transformers formats. Developed by LiamVisionary, it is optimized for private agent workflows and local experimentation. The model features a 32768 token context length and is specifically designed for use with tools like LM Studio and llama.cpp.

Loading preview...

Swarm Sovereign 12B: A Gemma 4-Based Chat Model for Local Inference

Swarm Sovereign 12B, developed by LiamVisionary, is an instruction-tuned chat model built upon the google/gemma-4-12B-it architecture. This 12 billion parameter model is specifically prepared for efficient local inference, supporting both GGUF and Hugging Face Transformers formats.

Key Capabilities & Features

  • Base Architecture: Utilizes the advanced Gemma 4 12B IT model from Google.
  • Local Inference Optimization: Provided in a Q4_K_M GGUF build (approx. 6.9 GB) for llama.cpp and LM Studio, and a Hugging Face safetensors build (approx. 22 GB) for Transformers-style loading.
  • Context Length: Supports a context window of 32768 tokens.
  • Identity: Responds as "Swarm Sovereign" in chat interactions.
  • Memory Efficiency: Estimated GPU memory usage of 7.71 GiB for 4096 context with full GPU offload in LM Studio.

Use Cases & Differentiators

This model is primarily intended for local experimentation and private agent workflows. Its availability in optimized GGUF format makes it highly accessible for developers looking to run powerful chat capabilities on their own hardware. It stands out as a specialized option for those building and managing private swarms of agents, offering a robust foundation for such applications.

Finetunes
1 models