# Qwen3.5-122B-A10B GPU requirements

Qwen3.5-122B-A10B has 125B parameters. To serve it, plan for about 300 GB of GPU memory in 16-bit, 150 GB in 8-bit or 75 GB in 4-bit, including about 20% for the KV cache and activations.

It is a mixture of experts: 256 experts, 8 of them active for each token. All of them stay in GPU memory, so plan memory for all 125B parameters.

[Deploy it on Arrakis](/console/#/models?model=Qwen/Qwen3.5-122B-A10B).

## What it takes to serve it

| Precision | Memory and the smallest setup |
|---|---|
| 16-bit | 300 GB: 4 × [H100](/gpus/h100.html.md), from $6.60 per hour for all 4 |
| 8-bit | 150 GB: one [B200](/gpus/b200.html.md), from $6.79 per GPU-hour |
| 4-bit | 75 GB: one [H100](/gpus/h100.html.md) or [A100](/gpus/a100.html.md), H100 from $1.65 per GPU-hour |

## What it takes to fine-tune it

| Method | Memory and the smallest setup |
|---|---|
| QLoRA | 88 GB: one [RTX PRO 6000](/gpus/rtx-pro-6000.html.md), from $1.90 per GPU-hour |
| LoRA | 300 GB: 4 × [H100](/gpus/h100.html.md), from $6.60 per hour for all 4 |
| Full fine-tuning | 2,001 GB: 8 × [B300](/gpus/b300.html.md), from $36.80 per hour for all 8 |

## About the model

| | |
|---|---|
| Parameters | `125B` |
| Active for each token | `10B` |
| Experts | `256` |
| Experts for each token | `8` |
| Context window (tokens) | `262,144` |
| Published in | `BF16` |
| License | `apache-2.0` |
| Released | `2026-02-24` |
| Weights | [Qwen/Qwen3.5-122B-A10B](https://huggingface.co/Qwen/Qwen3.5-122B-A10B) |

## Serve it yourself

```
vllm serve Qwen/Qwen3.5-122B-A10B --tensor-parallel-size 4
```

## Questions

**How much GPU memory does Qwen3.5-122B-A10B need?** To serve it, about 300 GB in 16-bit, 150 GB in 8-bit, 75 GB in 4-bit, including about 20% for the KV cache and activations. To fine-tune it, about 88 GB with QLoRA, 300 GB with LoRA and 2,001 GB for full fine-tuning with Adam, before activations.

**Can I run Qwen3.5-122B-A10B on one GPU?** Yes, quantized. In 8-bit it needs about 150 GB, which fits on one B200. In 16-bit it needs 4 H100 GPUs.

**What does it cost to run Qwen3.5-122B-A10B?** In 16-bit it needs 4 H100 GPUs; in 4-bit, one H100 or A100. Each setup on this page shows the lowest price per GPU-hour on Arrakis right now, and you pay by the hour from a prepaid balance.

**How do I deploy Qwen3.5-122B-A10B on Arrakis?** Open Models in the Arrakis console, paste Qwen/Qwen3.5-122B-A10B, choose vLLM, SGLang or llama.cpp, and rent a GPU it fits on. You get an OpenAI-compatible endpoint with its own API key, and root SSH to the machine.


More models: [/models](/models.html.md).