LLMs

Large language models (LLMs), APIs, agents, and harnesses.

Models

• Hugging face
• Unsloth

Qwen3.8-27B

• Unsloth: Qwen3.8-27B
• GSQ-RCO: Qwen3.8-27B
• MiaAI-Lab: Qwen3.8-27B on one DGX Spark.
• MiaAI-Lab: Qwen3.8-27B for one 24GB GPU or one DGX Spark.

Qwen3.8-Flash-Next

• Unsloth: Qwen3.8-Flash-Next
• qwen38-flash-next-spark on one DGX Spark.

Deepseek v4 flash

• MiaAI Lab: DS4f on one DGX Spark.
• Entrpi: DS4f on spark on one DGX Spark.

Gemma4

• Gemma4-31B-QAT
• Gemma4-26B-A4B-QAT

Speech recognition and text-to-speech

• VoxCPM

Document processing

• docling

Biology

• evo2

Runtime

• FreeToken : Optimized for MoE models.
• Unsloth desktop
• Lemonade for AMD GPUs.
• Lucebox
• Club 3090 : recipes for 3090/4090/5090 owners.

llama.cpp

• llama.cpp GitHub repo
• beellama.cpp : llama.cpp fork supporting KVarN KV cache format.
• ik_llama.cpp : llama.cpp fork with new quants and improved performance for MoE models.

vLLM

• vllm GitHub repo
• vllm-radiance docker image for 2 R9700’s.
• spark-vllm-docker : vLLM docker images for DGX sparks.

SGLang

• SGLang GitHub repo
• SGLang Cookbook
• MiaAI-Lab recipes

MCP servers

Model Context Protocol (MCP)

• DuckDuckGo
• Julia

Agents and harnesses

• Deepseek harness
• pi coding agent
• Qwen code
• Codewhale