Run Qwen3.8 Flash locally
Qwen3.8 Flash is 125B parameters (6B active per token — a mixture-of-experts, so it still needs room for all 125B in memory). Here's the VRAM it needs at each quantization, the smallest GPU that fits, and how a self-hosted box compares with the cheapest verified API.
VRAM by quantization
| Quantization | VRAM needed | Fits on | Notes |
|---|---|---|---|
| int4 (Q4) | 75 GB | A100 (80 GB) ($1.19/hr) | smallest footprint, minor quality loss |
| int8 (Q8) | 150 GB | 2× H100 (160 GB) ($3.98/hr) | near-lossless |
| fp16 (full) | 300 GB | 4× H100 (320 GB) ($7.96/hr) | reference quality |
VRAM ≈ params × bytes/param × 1.2 (overhead). See the fullmethod and break-even calculator.
Self-host vs API
Cheapest API for Qwen3.8 Flash is $0.62 / 1M tokens(blended) via Alibaba. Running it yourself in int4 fits aA100 (80 GB) at $1.19/hr — about $868.7/month at 24/7. Those two lines cross at roughly 1.4B tokens/month: below that the API wins on cost, above it the dedicated GPU does (assuming you keep it busy). Tune your own volume in the break-even calculator.
Run it at home
No cloud account needed: Qwen3.8 Flash in int4 (75 GB) fits a Mac Studio M5 Ultra (96 GB unified), a $5,499 one-off buy. Amortized over 3 years that's about $0.2092/hr — hardware only, electricity aside. Against the Alibaba API at $0.62/1M, buying it pays for itself after roughly 8.9B tokens total. Below that the API is cheaper; a machine that mostly sits idle rarely earns back its price.
Get the weights
Quantized builds on the model hubs — links search live, so they track new builds as they appear:
- GGUF builds llama.cpp · Ollama · LM Studio
- AWQ builds GPU serving (vLLM)
- GPTQ builds GPU serving
- MLX builds Apple silicon
- Ollama library one-command local run
License: open weights under Apache 2.0 · terms. Compare all providers on the qwen3.8-flash API page.