# Qwen3-Coder-Next GPU requirements

Qwen3-Coder-Next has 80B parameters. To serve it, plan for about 191 GB of GPU memory in 16-bit, 96 GB in 8-bit or 48 GB in 4-bit, including about 20% for the KV cache and activations.

It is a mixture of experts: 512 experts, 10 of them active for each token. All of them stay in GPU memory, so plan memory for all 80B parameters.

[Deploy it on Arrakis](/console/#/models?model=Qwen/Qwen3-Coder-Next).

## What it takes to serve it

| Precision | Memory and the smallest setup |
|---|---|
| 16-bit | 191 GB: one [B300](/gpus/b300.html.md), from $4.60 per GPU-hour |
| 8-bit | 96 GB: one [RTX PRO 6000](/gpus/rtx-pro-6000.html.md), from $1.90 per GPU-hour |
| 4-bit | 48 GB: one [L40S](/gpus/l40s.html.md), [RTX 6000 Ada](/gpus/rtx-6000-ada.html.md) or [RTX A6000](/gpus/rtx-a6000.html.md), L40S from $0.44 per GPU-hour |

## What it takes to fine-tune it

| Method | Memory and the smallest setup |
|---|---|
| QLoRA | 56 GB: one [H100](/gpus/h100.html.md) or [A100](/gpus/a100.html.md), H100 from $1.65 per GPU-hour |
| LoRA | 191 GB: one [B300](/gpus/b300.html.md), from $4.60 per GPU-hour |
| Full fine-tuning | 1,275 GB: 8 × [B200](/gpus/b200.html.md), from $54.32 per hour for all 8 |

## About the model

| | |
|---|---|
| Parameters | `80B` |
| Experts | `512` |
| Experts for each token | `10` |
| Context window (tokens) | `262,144` |
| Published in | `BF16` |
| License | `apache-2.0` |
| Released | `2026-01-30` |
| Weights | [Qwen/Qwen3-Coder-Next](https://huggingface.co/Qwen/Qwen3-Coder-Next) |

## Serve it yourself

```
vllm serve Qwen/Qwen3-Coder-Next --tensor-parallel-size 1
```

## Questions

**How much GPU memory does Qwen3-Coder-Next need?** To serve it, about 191 GB in 16-bit, 96 GB in 8-bit, 48 GB in 4-bit, including about 20% for the KV cache and activations. To fine-tune it, about 56 GB with QLoRA, 191 GB with LoRA and 1,275 GB for full fine-tuning with Adam, before activations.

**Can I run Qwen3-Coder-Next on one GPU?** Yes. In 16-bit it needs about 191 GB, which fits on one B300.

**What does it cost to run Qwen3-Coder-Next?** In 16-bit it needs one B300; in 4-bit, one L40S, RTX 6000 Ada or RTX A6000. 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-Coder-Next on Arrakis?** Open Models in the Arrakis console, paste Qwen/Qwen3-Coder-Next, 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).