Qwen3.8 Flash Next 125B-A6B
By Qwen · China
Updated 2026-08-28
Overview
Qwen3.8 Flash Next is a multimodal MoE model from Qwen with 125B total parameters and 6B active per token, handling vision, code, and chat with 256K context — self-hostable on a single 80GB GPU at Q4.
When to pick this model
- Self-hosted multimodal chat combining vision, code, and text
- Long-document or long-codebase analysis at 256K context
- Teams with a single 80GB-class GPU wanting high throughput per dollar
- Workloads that need MoE-level speed without full dense-125B cost
VRAM requirements by quantization
| Quantization | VRAM required |
|---|---|
| Q4_K_M (recommended) | 72 GB |
| Q5_K_M | 89 GB |
| Q8_0 | 134 GB |
| FP16 (no quantization) | 250 GB |
VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.
In practice, Qwen3.8 Flash Next 125B-A6B is server-class even at Q4_K_M (72 GB). Stepping up to Q8_0 nearly doubles the footprint to 134 GB, and unquantized FP16 weights take 250 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.
Without a GPU, Qwen3.8 Flash Next 125B-A6B needs roughly 162 GB of system RAM to run on CPU via llama.cpp or Ollama — workable for background jobs, but far slower than GPU inference. Throughput estimates from our compatibility engine: around 32 tokens/sec on entry-level GPUs, on the order of 50 tokens/sec on a mid-range card, and up to 75 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.
What hardware do you need
The table below matches Qwen3.8 Flash Next 125B-A6B to common GPU memory tiers, using the highest-fidelity quantization that fully fits each card class. Spilling layers to system RAM works but costs most of the speed, so size your card to the quantization you actually want to run.
| GPU memory | Example cards | Best fit for Qwen3.8 Flash Next 125B-A6B |
|---|---|---|
| 8 GB | RTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GB | Does not fit — needs 72 GB at Q4_K_M |
| 12 GB | RTX 5070, RTX 5070 Ti Laptop, RTX 4080 Laptop | Does not fit — needs 72 GB at Q4_K_M |
| 16 GB | RTX 5080, RTX 4080 Super, Radeon RX 9070 XT | Does not fit — needs 72 GB at Q4_K_M |
| 24 GB | RTX 4090, Radeon RX 7900 XTX, RTX 5090 Laptop | Does not fit — needs 72 GB at Q4_K_M |
| 32 GB | RTX 5090 | Does not fit — needs 72 GB at Q4_K_M |
Which GPU should you buy to run Qwen3.8 Flash Next 125B-A6B?
To run Qwen3.8 Flash Next 125B-A6B locally at Q4, you need ~72 GB of VRAM. The best value for this is a Apple Mac Studio (64+ GB unified memory).
As an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. It does not influence our independent rankings.
Strengths
- MoE design (125B total / 6B active) delivers high throughput for its size
- Native 256K context
- Multimodal — vision, code, and chat in one model
- Fits a single 80GB GPU at Q4 (~72GB)
Limitations
- Still server-class hardware — ~72GB VRAM at Q4
- "Other" license — check usage terms before commercial deployment
- Recent release — quantized ecosystem and tooling still maturing
Typical workloads
In our catalog grid, Qwen3.8 Flash Next 125B-A6B is filed under Multimodal Chat & Code, Long Context 256K, Single-GPU Server — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: code generation and review (pair it with an editor integration like Continue.dev or Cline); vision-language work — screenshots, charts, scanned documents; multilingual workloads.
The 250k-token context window is large enough to hold entire codebases' worth of files or long reports in a single prompt, which is what makes local RAG and document analysis practical. It ships under the Autre (open weights) license — commercial use is generally possible but read the specific terms before embedding it in a product.
Architecture & training
Architecture: Multimodal MoE · 125B total parameters / 6B active per token · 256K context
Training: Flash Next model in the Qwen3.8 family, multimodal (vision + text) Mixture-of-Experts architecture. 'Other' license (open weights).
A strong self-hosted multimodal option if you have an 80GB GPU and want MoE speed with long context.
Quick start
ollama pull qwen3.8-flash-nextOr use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.
Similar models worth comparing
Frequently asked questions
How much VRAM does Qwen3.8 Flash Next 125B-A6B need?
At the recommended Q4_K_M quantization, Qwen3.8 Flash Next 125B-A6B needs about 72 GB of VRAM. Q8_0 takes 134 GB, and unquantized FP16 weights take 250 GB.
Can Qwen3.8 Flash Next 125B-A6B run without a GPU?
Yes — with roughly 162 GB of system RAM it runs CPU-only through llama.cpp or Ollama. Expect a fraction of GPU speed, which is fine for background or batch jobs but slow for interactive chat.
What context window does Qwen3.8 Flash Next 125B-A6B support?
Qwen3.8 Flash Next 125B-A6B supports a 250k-token context window (256,000 tokens).
Can I use Qwen3.8 Flash Next 125B-A6B commercially?
Qwen3.8 Flash Next 125B-A6B ships under the Autre (open weights) license. Commercial use is generally permitted subject to its terms — review the license text before shipping a product.
How fast is Qwen3.8 Flash Next 125B-A6B on consumer hardware?
Our compatibility engine estimates on the order of 50 tokens/sec on a mid-range GPU and up to 75 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.
Which quantization of Qwen3.8 Flash Next 125B-A6B should I download first?
Start with Q4_K_M (72 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. It does not fit a single 24 GB consumer card — plan for multi-GPU or server hardware.
Is Qwen3.8 Flash Next 125B-A6B the right pick for you?