
outsourc-e/Qwen3.8-27B-Unleashed-GGUF · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
This curated guide brings together 13 vetted Mixture-of-Experts resources, including 13 with structured technical analysis and 0 with practitioner field notes. Primary focus areas include LLM Inference Engine PR, LLM Inference Server, LLM Inference Tool, LLM Model Card, Local LLM Inference Engine. Primary sources include github.com, huggingface.co, deepgrove.ai, dwarfstar.sh, perplexity.ai.
| Resource | Category & Format | Key Takeaway & Notes |
|---|---|---|
| ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF · Hugging Facehuggingface.co | ML Model Repository / Model repository with documentation / Free download from Hugging Face | Per-tensor non-uniform quantization assigns a separate quantization type to every tensor via gradient-based search under a size budget |
| GitHub - Niko1221/Strata: Qwen3.8-Flash-Next (125B MoE) on a 8GB+ NVIDIA GPU: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.github.com | Local LLM Inference Tool / open-source software project with documentation / free and open source on GitHub | Runs a 125B-parameter MoE model on a single consumer GPU (12-24 GB VRAM) with 64 GB system RAM, eliminating the need for server-grade hardware |
| DwarfStar 4 (ds4): Local DeepSeek V4.1, Qwen and GLMdwarfstar.sh | Local LLM Inference Engine / project documentation and website / open source, MIT license | Asymmetric 2-bit quantization compresses routed MoE experts while preserving critical shared paths for accuracy |
| GitHub - claudejaune/Qwen3.8-Halogen-Flash-Strix-Halo: Run Qwen3.8-Flash-Next (180B, 6B active) on a single AMD Strix Halo via the halogen-flash-server containergithub.com | Local LLM Inference Setup / GitHub repository with guided shell scripts / Open source, free; requires AMD Strix Halo hardware with 128 GB RAM | Runs a 180B-parameter MoE model (6B active) on a single AMD Strix Halo with 128 GB RAM — no GPU cluster needed |
| GitHub - julianmb/halofpx: High-performance Lemonade alternative for AMD Strix Halo & Radeon — latest MoE models (Ling-3.0-Flash, Qwen 3.8 Flash Next, Ornith 1.5), bleeding-edge RDNA 3.5 Wave64/ROCmFP4 kernels, and silicon-tuned profiles.github.com | LLM Inference Server / Open-source software project / Open source on GitHub | Custom Mesa RADV Wave64 cooperative matrix kernels and ROCmFP4 quantization deliver +13.5% decode speed and -16.7% model size vs standard GGUF |
| Optimizing On-Device Inference for Apple Siliconperplexity.ai | ML Inference Engineering / technical blog post / free online | Rust runtime with custom Metal kernels eliminates PyTorch and MLX from the execution path entirely |
| GitHub - quimmedes/cafe-llama.cpp: A experimental fork of llama.cpp with LLM inference in C/C++github.com | LLM Inference Tool / Open source GitHub repository with documentation / Open source, free on GitHub | Provides flags to offload MoE expert weights to pinned host RAM (zero-copy DMA) or CPU RAM while keeping attention and KV cache on GPU |
| feat(qwen4_exp): support Qwen3.8-Flash-Next by jason-fxz · Pull Request #257 · FlashML-org/FreeTokengithub.com | LLM Inference Engine PR / GitHub Pull Request / open source | Hybrid architecture: 36 gated delta net layers + 12 compressed block-sparse attention layers on 4 hyper-connection residual streams |
| GitHub - FlashML-org/FreeTokengithub.com | Machine Learning Software / open source software / open source | Enables running 290B+ parameter MoE models locally on consumer GPUs (NVIDIA RTX 30, 40, and 50 series). |
| Maple-Preview Model Card — DeepGrovedeepgrove.ai | LLM Model Card / model card / technical blog post / open source, weights on Hugging Face, chat demo online | SOTA in its weight class: 87.5 on AIME 26 and 73.5 on GPQA-D, competitive with larger models |
| JetBrains/Mellum2-12B-A2.5B-Thinking — 12B / 2.5B active · MOE · 128K ctxrecipes.vllm.ai | Machine Learning Model / technical guide / open source | Supports a 131,072-token context window using a hybrid of sliding-window and full-attention layers. |
| NVIDIA Nemotron-3-Super: How To Run Guide | Unsloth Documentationunsloth.ai | Technical Documentation / technical guide / free online | Supports local inference on devices with 64GB RAM/VRAM and fine-tuning via Unsloth. |
| unsloth/Kimi-K2-Instruct-GGUF · Hugging Facehuggingface.co | Machine Learning Model Repository / model checkpoint in GGUF format with documentation / open source download from Hugging Face | Mixture‑of‑Experts LLM with 1 trillion total parameters and 32 B active parameters |
A collection of bookmarks filtered by the tag "Mixture-of-Experts".
Showing 1-13 of 13 bookmarks tagged with "Mixture-of-Experts"
All Stories