Open weights · Qwen Community 1.0
Qwen3.8-Flash-Next
Alibaba's open-weight preview of the architecture it says will underpin Qwen4: a 125B mixture-of-experts model with 6B active per token, under the Qwen Community 1.0 licence.
Key Specifications
- Context window
- 262K tokens
- Max output
- 131K tokens
- Released
- Parameters
- 125B, 6B active
- Inputs
- Text, image, video
- Capabilities Show details
- Reasoning Thinks before it answers, always on or as a switchable mode. Function calling Connect to external tools, APIs, and systems. Structured output Return responses in structured formats like JSON.
Hosted API pricing
| Provider | Input / 1M tokens | Output / 1M tokens | Cost at 10M in + 2M out | |
|---|---|---|---|---|
|
Alibaba Cloud
Creator
|
Open weights, no hosted price | View | ||
Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.
Estimated cost to self-host
Qwen3.8-Flash-Next needs about 72 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× RTX 3060 at about $288 a month.
| Precision | Cheapest, 32K context | Cheapest, full 262K context |
|---|---|---|
| 4-bitINT4 / FP4 |
8× RTX 3060
72 GB VRAM · $288/mo
|
8× RTX 3060
77 GB VRAM · $288/mo
|
| 8-bitFP8 / INT8 |
8× RTX 3090
128 GB VRAM · $806/mo
|
8× RTX 3090
133 GB VRAM · $806/mo
|
| 16-bitFP16 / BF16 |
8× RTX A6000
253 GB VRAM · $3,168/mo
|
8× RTX A6000
258 GB VRAM · $3,168/mo
|
Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. No guarantee of runtime support, usable performance, or that a matching quantized build exists. How we estimate costs.
Frequently Asked Questions
What is Qwen3.8-Flash-Next good for?
Text, image and video agent and coding work you run yourself, with thinking on by default and a reasoning effort setting. Qwen Community 1.0 weights, on a multi-GPU node.
When is Qwen3.8-Flash-Next not a good fit?
Audio input or image output, and there is no built-in web search, though a host or your own tool loop can add one. It takes a multi-GPU node, and Alibaba calls it an experimental preview, with the hosted Qwen3.8-Flash carrying the production features instead.
What is the cheapest way to run Qwen3.8-Flash-Next?
No provider we track hosts it; self-hosting on the cheapest rental that fits (8x RTX 3060, $0.40 per hour) costs about $288 a month.
Can I self-host Qwen3.8-Flash-Next?
Yes. The weights are Qwen Community 1.0 licensed. At 4-bit it needs about 72 GB of GPU memory, which starts at roughly $288 a month on the cheapest rental that fits.
More from Alibaba Cloud
| Model | Context | Input / 1M | Output / 1M |
|---|---|---|---|
|
Qwen3-Coder-Plus
|
1M | $1.00 | $5.00 |
|
Qwen3.5-Flash
|
1M | $0.10 | $0.40 |
|
Qwen3.5-Plus
|
1M | $0.40 | $2.40 |
|
Qwen3.6-Flash
|
1M | $0.25 | $1.50 |
|
Qwen3.6-Plus
|
1M | $0.50 | $3.00 |
|
Qwen3.7-Flash
|
1M | $0.03 | $0.13 |
|
Qwen3.7-Max
|
1M | $1.25 | $3.75 |
|
Qwen3.7-Plus
|
1M | $0.28 | $1.10 |