Reasoning
How to deploy Qwen3-235B-A22B-Thinking-2507 on a GPU cloud
A 235B (MoE) model tuned to reason step by step before answering. Full specs, license and use cases.
Qwen3-235B-A22B-Thinking-2507 size and hardware requirements
235.1B
Total parameters
~22B active per token (mixture-of-experts; see total parameters above)
Active parameters
BF16
Published precision
525.5 GB
Min VRAM (native)
| Precision | Weight size on disk | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 437.9 GB | 525.5 GB | A100 | 7 | $5.95/hr |
| FP8 (quantized) | 218.9 GB | 262.7 GB | RTX 6000 Ada | 6 | $3.14/hr |
| INT4 (quantized) | 109.5 GB | 131.4 GB | RTX 3090 | 6 | $0.882/hr |
How to run Qwen3-235B-A22B-Thinking-2507
Run Qwen3-235B-A22B-Thinking-2507 with vLLM
From Qwen/Qwen3-235B-A22B-Thinking-2507's own deployment docs.
vllm serve Qwen/Qwen3-235B-A22B-Thinking-2507 --tensor-parallel-size 8 --max-model-len 262144 --enable-reasoning --reasoning-parser deepseek_r1Source: https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507/raw/main/README.md
Run Qwen3-235B-A22B-Thinking-2507 with GGUF quantizations
Prebuilt GGUF weights published at unsloth/Qwen3-235B-A22B-Thinking-2507-GGUF. Run with llama.cpp's llama-server or load the repo directly in LM Studio.
llama-server -hf unsloth/Qwen3-235B-A22B-Thinking-2507-GGUFSource: https://huggingface.co/unsloth/Qwen3-235B-A22B-Thinking-2507-GGUF
Deploy Qwen3-235B-A22B-Thinking-2507 on Aquanode
Aquanode has no one-click deploy template for Qwen3-235B-A22B-Thinking-2507; you install the inference engine yourself with the commands below. Aquanode sells GPU pods billed per second, not a hosted inference API.
- Launch a bare GPU pod sized to the requirement above (7× A100 or larger).
- Open a terminal on the pod, or save one of the commands above as a startup script so it runs automatically the first time the pod boots.
- Run the command and connect to the resulting endpoint.
Launch a GPU pod