How to Deploy Qwen3.8 Flash Next Locally
Qwen3.8 Flash Next is a MoE model with 176B total parameters and 6B active per token. At Q4 it needs 100GB of VRAM to sit entirely on the GPU; with experts offloaded to system RAM it needs only 4GB of VRAM plus 97GB of RAM. The cheapest card that runs it today is the Ryzen AI Max+ 395 128GB at about $3,913, at roughly 31.8 tokens/s. In offload mode the cheapest card that runs it is the Tesla V100 16GB, at roughly 22 tokens/s.
Qwen3.8 Flash Next details
How much VRAM does Qwen3.8 Flash Next need for local deployment?
MoE: all 176B parameters must stay resident in memory, while speed is set by the 6B active per token.
| Quantization | WeightsGB | Total at 8KGB | Total at 32KGB | Total at 128KGB |
|---|---|---|---|---|
| 1-bit | 74.5GGUF | 42.5 | 43.1 | 45.3 |
| 2-bit | 78.9GGUF | 66.6 | 67.2 | 69.4 |
| 3-bit | 90GGUF | 85.5 | 86.1 | 88.3 |
| 4-bit | 111.3GGUF | 99.3 | 99.9 | 102.1 |
| 5-bit | 158.3GGUF | 119.9 | 120.5 | 122.8 |
| 8-bit | 188.2GGUF | 183.6 | 184.2 | 186.4 |
| FP16 | 354GGUF | 345.4 | 346 | 348.2 |
Total = weights + KV cache (FP16) + 1GB runtime overhead. Weights marked GGUF are measured file sizes from the repository; the rest are estimated from bytes per parameter. Contexts beyond this model's 256K limit show a dash. Each tier takes whichever file of that bit width the repository has, preferring unsloth dynamic quants, so the exact file name varies by model.
Which GPUs can run Qwen3.8 Flash Next?
AmazonJP
| Device | VRAMGB | Price | Est. t/s | Suggested context |
|---|---|---|---|---|
 Ryzen AI Max+ 395 128GB | 128 | $3,9136 listings | 31.8 | 128K |
 DGX Spark 128GB | 128 | $8,2233 listings | 44.7 | 128K |
| Mac Studio M4 Max 128GB | 128 | —No listings | 63.7 | 128K |
| Mac Studio M5 Max 128GB | 128 | —No listings | 69.4 | 128K |
| MacBook Pro M4 Max 128GB | 128 | —No listings | 63.7 | 128K |
| MacBook Pro M5 Max 128GB | 128 | —No listings | 69.4 | 128K |
| Mac Studio M3 Ultra 512GB | 512 | —No listings | 84.7 | 128K |
Fits when weights + KV cache + runtime overhead is at or below usable memory, single card. "Offload" on MoE models means it runs with expert weights in system RAM (only attention layers and KV cache stay in VRAM); that speed assumes 70 GB/s RAM bandwidth. The context column is the largest tier that fits.
Recommended GPUs for Qwen3.8 Flash Next with CPU/GPU offloading
AmazonJP
Expert weights live in system RAM (needs at least 97 GB); VRAM holds only attention layers, shared experts and the KV cache, at least 4 GB at Q4. Speed is bound by RAM bandwidth (estimated at 70 GB/s) and is far slower than a full-VRAM setup.
Deploy Qwen3.8 Flash Next locally with Ollama, llama.cpp or vLLM
The Ollama library has no official tag for it yet.
llama.cpp pulls the GGUF straight from Hugging Face; -c sets the context length.
llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q4_K_M -c 8192
vLLM serves the original-precision weights, which needs far more memory than GGUF and usually more than one GPU.
vllm serve Qwen/Qwen3.8-Flash-Next
File names follow that week's model-library snapshot; the definitions are in the methodology. Methodology
FAQ
How much VRAM does Qwen3.8 Flash Next need?
At Q4_K_M with an 8K context Qwen3.8 Flash Next needs about 100GB of VRAM; 120GB leaves comfortable headroom.
What is the cheapest GPU that runs Qwen3.8 Flash Next?
The Ryzen AI Max+ 395 128GB: 128GB of VRAM, currently about $3,913, at roughly 31.8 tokens/s.
Can you run Qwen3.8 Flash Next with Ollama?
Not yet: the Ollama library has no official tag for it. You can load the GGUF with llama.cpp instead.