AIQB
TutorialsOrdinary

Flash-MoE: How to Run a 397B Model on a Laptop

Source: DEV Community·
Qwen

Summary

Flash-MoE streams a 397B Qwen model from SSD on a 48GB MacBook at 5.5 tok/s. No cloud. Here's how sparse MoE activation makes this actually work.
CategoryAI Tutorials & Practice
TierOrdinary
Published
Indexed by AIQB
SourceDEV Community
AIQB record IDintel-48470e4def94b9e7e72ea923
Read original source ↗More Tutorials

Related stories

Matched by category, entity tags and source

TutorialsOrdinary
Qwen 3.6 Plus vs Claude Opus 4.6: Which One Actually Works in Production?
Qwen 3.6 Plus is free, fast, and ships 1M context. Claude Opus 4.6 costs more but handles hard agentic tasks better. Here's what actually matters.
Source: DEV Community · Qwen / Claude
TutorialsOrdinary
I fine-tuned Qwen3-14B for my AI companion app for $1.30. The hard part wasn't the model
Soulor AI is an AI companion app with persistent memory and five relationship stages, from Stranger...
Source: DEV Community · Qwen
TutorialsOrdinary
Export ChatGPT, Claude, Gemini, Grok, DeepSeek, Kimi, and Qwen chats to Markdown with Convolvger 0.6.0
If you have ever tried to export a ChatGPT conversation to Markdown, or save a Claude chat as a .md...
Source: DEV Community · ChatGPT / DeepSeek / Qwen
TutorialsOrdinary
The provider that failed was the feature: building a model-agnostic AI arbiter (Kimi K2 · Qwen · EIP-712)
A real failure story from building EscrowLens, a passkey-native escrow protocol on Monad for the...
Source: DEV Community · Qwen / Kimi
TutorialsOrdinary
Claude Code vs Qwen Code vs Codex
AI coding tools have changed. A couple of years ago, we were excited when autocomplete correctly...
Source: DEV Community · Qwen / Claude
TutorialsOrdinary
Qwen3.8-27B: An Architecture and Hands-On Look at Alibaba's Compact Dense Model
In August 2026, less than a month after Moonshot AI's Kimi-K3 surprised the world, Alibaba Group...
Source: DEV Community · Qwen / Kimi
TutorialsOrdinary
Hexagonal Architecture for Provider-Agnostic RAG Pipelines
Demo stack: LangChain, Llama.cpp, Mistral.ai, Qwen Embeddings, PostgreSQL, Telegraf, Prometheus on...
Source: DEV Community · Qwen / Llama / Mistral
TutorialsOrdinary
I Rebuilt Jev's Structure with Qwen (Not Its Capabilities)
Jev does not generate an answer string. It makes a typed decision in one forward pass. So,...
Source: DEV Community · Qwen
TutorialsOrdinary
Qianwen Office generates a multi-user app from a prompt. The prompt is the easy part.
On September 7, Alibaba's Qianwen Office (千问办公) launched a multi-person workspace. You describe who...
Source: DEV Community · Qwen
TutorialsOrdinary
Best Local LLM for Coding: 8GB to 24GB VRAM Picks
The best local LLM for coding by VRAM bracket: Qwen3 Coder vs DeepSeek at 8, 12, 16 and 24 GB, the right quant per card, plus a runnable Ollama setup.
Source: DEV Community · DeepSeek / Qwen / Ollama