Qwen3 8B
Pilot $4,900 fixed
- Sizes
- 8B
- Active params
- All parameters (dense)
- Context
- 32K (128K with YaRN)
- Licence
- Apache 2.0
- Fits on
- CPU only at 4-bit, 1× T4 16GB at FP8
- Endpoint
- llm.yourcompany.internal/v1
- Egress
- 0
- Owner
- YOU
Qwen3on your own GPUs.
Qwen3 in three sizes, served from GPUs you control. Apache 2.0 licence, 32K native context (128K with YaRN scaling), nothing leaves your network.
This card is the installation. Pick a size on its edge and it is re-issued.
The 72-hour pilot with Qwen3.
$4,900 fixed. The 32B model is the usual starting point on one 48GB card at FP8. The 235B-A22B model activates 22B parameters per token and needs 2× H100 (160GB) at 4-bit, 4× H100 (320GB) at FP8.
Live on your data
72 h$4,900 fixed
Qwen3 live on your data in 72 hours. The clock starts after access and prerequisites.
Your endpoint
/v1OpenAI-compatible
OpenAI-compatible endpoint. Your applications switch by changing the base URL.
Your hardware
- Qwen3 8BCPU only at 4-bit, 1× T4 16GB at FP8
- Qwen3 32B1× RTX 4090 24GB at 4-bit, 1× A100 40GB at FP8
- Qwen3 235B-A22B2× H100 (160GB) at 4-bit, 4× H100 (320GB) at FP8
The readout
Evalon your data
Reference evaluation on your data. Readout and production scoping.
We size the hardware and source it at cost, with no markup. Ballpark: one H100 80GB ≈ $25–30k; a 4×H100 node ≈ $180–240k; smaller models run on far less.
Same SDK. Change the base URL.
The model is one parameter. Everything else in your application stays as it is.
client = OpenAI(
base_url="https://llm.yourcompany.internal/v1", # inside your network
api_key="your-key",
)
reply = client.chat.completions.create(
model="qwen3-8b",
messages=[{"role": "user", "content": "Summarise this contract."}],
)Licence, plainly.
- Qwen3 8BApache 2.0
- Qwen3 32BApache 2.0
- Qwen3 235B-A22BApache 2.0
Qwen3 is released under Apache 2.0. Commercial use, modification and private deployment are permitted, and you keep your fine-tuned weights.
Newer Qwen generations (Qwen3.5, 3.6 and 3.8) are in the catalogue too. Most are Apache 2.0; Qwen3.8-Flash-Next carries the Qwen Community License with a revenue clause.
Published by Alibaba. We are not affiliated with the vendor. This is a summary, not legal advice.
Related models.
- Qwen3.8Alibaba2 models, from 1× RTX 4090 24GB at 4-bitApache 2.0 / Qwen Community License 1.0 (revenue cap)
- Gemma 4Google2 models, from 1× T4 16GB at 4-bitApache 2.0
- gpt-ossOpenAI2 models, from 1× T4 16GB at 4-bitApache 2.0
Talk to an engineer.
Tell us the model, the data and the GPUs you have. An engineer from Limassol or Auckland scopes the 72-hour pilot with you.