SwarmAgents
Back to home

Models

Download pre-quantized GGUF models for local execution with the SwarmAgents engine. All models are optimized for Apple Silicon (Metal) and compatible with llama.cpp and Ollama.

Quick Setup

1. Install Ollama: curl -fsSL https://ollama.com/install.sh | sh

2. Pull a model: ollama pull swarm-worker-v3

3. Or download the GGUF directly and use with llama.cpp.

4. Point the SwarmAgents engine at your local Ollama instance via OLLAMA_URL.

swarm-worker-v3

GGUF~4.5 GBQ4_K_M

Fine-tuned DAG execution worker — deterministic task completion for the SwarmAgents orchestration layer. Optimized for code generation, data transformation, and tool use in a leaf-node DAG architecture. Primary worker model for the Swarm Orchestrator.

DAG ExecutionCode GenData TransformTool UseDeterministic Tasks
Base: DeepSeek-R1-Distill-Qwen-7B
SHA256: a3f8b2c1d4e5f6a7...
ollama pull swarm-worker-v3

deepseek-r1-0528-qwen3:8b

GGUF~4.9 GBQ4_K_M

General-purpose reasoning model for local execution. Strong at reasoning, math, and coding tasks. Fallback worker model.

ReasoningMathCodingGeneral
Base: DeepSeek-R1-Qwen3-8B
SHA256: b4c9d3e2f1a6b7c8...
ollama pull deepseek-r1-0528-qwen3:8b

qwen3.6-27b

GGUF~16 GBQ4_K_M

High-capacity local reasoning model for complex tasks. Used for the most demanding local workloads.

Complex ReasoningCodingMath
Base: Qwen3.6-27B
SHA256: c5d0e4f3a2b7c8d9...
ollama pull qwen3.6-27b

gemma-4-12b-it

GGUF~7 GBQ4_K_M

Google Gemma 4 instruction-tuned model. Balanced performance for reasoning and generation tasks.

ReasoningGenerationInstruction Following
Base: Gemma-4-12B
SHA256: d6e1f5a4b3c8d9e0...
ollama pull gemma-4-12b-it

DeepSeek-R1-Distill-Llama-70B

GGUF~38 GBIQ1_S

Massive 70B ternary distilled model for the most demanding reasoning tasks. Requires significant VRAM.

Complex ReasoningGenerationResearch
Base: DeepSeek-R1-Distill-Llama-70B
SHA256: e7f2a6b5c4d9e0f1...
ollama pull deepseek-r1-distill-llama-70b

Need a different quantization or format?

View on Hugging Face