# Qwen3-Coder-30B-A3B-Instruct

> OnPrem World · https://onpremworld.com/en/models/qwen3-coder-30b-a3b · Last updated: 2026-10-14 · CC BY 4.0

The coding version of Qwen3-30B-A3B (3.3B active): a fast local coding assistant on a 20–24 GB card, with a 256K context and fill-in-the-middle completion.

|  |  |
|---|---|
| Developer | Alibaba Qwen |
| Parameters | 30.5B (3.3B active, MoE) |
| Native context | 262,144 tokens |
| Licence | Apache-2.0 |
| Commercial use | Yes |
| KV cache | All 48 layers (standard attention) |
| Released | 2025-07 |

```
ollama run qwen3-coder:30b
```

## Memory needed

Weights + KV cache + runtime overhead, in GiB. Lower quantization = less memory, slightly lower quality.

| Quantization | 4K | 8K | 32K |
|---|---|---|---|
| Q4_K_M | 19.0 GiB | 19.3 GiB | 21.6 GiB |
| Q5_K_M | 22.1 GiB | 22.5 GiB | 24.7 GiB |
| Q6_K | 25.4 GiB | 25.8 GiB | 28.1 GiB |
| Q8_0 | 32.6 GiB | 32.9 GiB | 35.2 GiB |
| FP16 | 60.5 GiB | 60.9 GiB | 63.2 GiB |

## Which hardware runs it (8K)

| Hardware | Memory | Best quant | tok/s (est.) |
|---|---|---|---|
| GeForce RTX 5090 | 32 GB | Q6_K | 461 |
| GeForce RTX 4090 | 24 GB | Q5_K_M | 301 |
| Radeon RX 7900 XTX | 24 GB | Q5_K_M | 286 |
| Radeon RX 7900 XT | 20 GB | Q4_K_M | 280 |
| GeForce RTX 3090 | 24 GB | Q5_K_M | 279 |
| RTX 6000 Ada Generation | 48 GB | Q8_0 | 192 |
| RTX PRO 6000 Blackwell Workstation Edition | 96 GB | FP16 | 190 |
| Radeon AI PRO R9700 | 32 GB | Q6_K | 165 |
| Arc Pro B70 | 32 GB | Q6_K | 157 |
| RTX A6000 | 48 GB | Q8_0 | 153 |
| RTX PRO 5000 72GB Blackwell | 72 GB | FP16 | 143 |
| Arc Pro B60 | 24 GB | Q5_K_M | 136 |
| M5 Ultra | 512 GB | FP16 | 118 |
| M5 Max (18-core CPU / 32-core GPU) | 36 GB | Q6_K | 110 |
| M4 Max (14-core CPU / 32-core GPU) | 36 GB | Q6_K | 98 |
| M3 Ultra | 512 GB | FP16 | 81 |
| M5 Max (18-core CPU / 40-core GPU) | 128 GB | FP16 | 60 |
| M5 Pro | 64 GB | Q8_0 | 57 |
| M4 Max (16-core CPU / 40-core GPU) | 128 GB | FP16 | 54 |
| M4 Pro | 64 GB | Q8_0 | 51 |
| M6 (24GB/32GB configurations, 170 GB/s) | 32 GB | Q5_K_M | 47 |
| M5 | 32 GB | Q5_K_M | 42 |
| RTX Spark (N1X) laptops & compact desktops | 128 GB | FP16 | 27 |
| DGX Spark | 128 GB | FP16 | 25 |
| Ryzen AI Max+ PRO 495 (Gorgon Halo) systems | 192 GB | FP16 | 25 |
| Ryzen AI Max+ 395 (Strix Halo) systems | 128 GB | FP16 | 23 |
| GeForce RTX 3060 (12 GB) | 12 GB | Doesn't fit | – |
| GeForce RTX 4060 Ti 16GB | 16 GB | Doesn't fit | – |
| GeForce RTX 4080 SUPER | 16 GB | Doesn't fit | – |
| GeForce RTX 5060 Ti 16GB | 16 GB | Doesn't fit | – |
| GeForce RTX 5070 | 12 GB | Doesn't fit | – |
| GeForce RTX 5070 Ti | 16 GB | Doesn't fit | – |
| GeForce RTX 5080 | 16 GB | Doesn't fit | – |
| Radeon RX 9060 XT 16GB | 16 GB | Doesn't fit | – |
| Radeon RX 9070 XT | 16 GB | Doesn't fit | – |
| Arc B580 | 12 GB | Doesn't fit | – |
| M6 (16GB configuration, 153 GB/s) | 16 GB | Doesn't fit | – |

## Frequently asked questions

### How much VRAM does Qwen3-Coder-30B-A3B-Instruct need?

With an 8K context, about 19.3 GiB at Q4_K_M and 32.9 GiB at Q8_0 (weights + KV cache + runtime overhead). Longer contexts need more memory for the KV cache.

### Can Qwen3-Coder-30B-A3B-Instruct run on an GeForce RTX 4090?

Yes. Qwen3-Coder-30B-A3B-Instruct fits on the GeForce RTX 4090 (24 GB) at Q5_K_M with an 8K context, using about 22.5 GiB, at an estimated 301 tokens/s.

### Can I run Qwen3-Coder-30B-A3B-Instruct on a Mac?

Yes. Qwen3-Coder-30B-A3B-Instruct fits on the M3 Ultra (96 GB) at FP16 with an 8K context, using about 60.9 GiB, at an estimated 81 tokens/s.

### Can Qwen3-Coder-30B-A3B-Instruct be used commercially?

Qwen3-Coder-30B-A3B-Instruct is released under Apache-2.0. The licence allows commercial use (read its conditions).

## Sources

- https://simonwillison.net/2025/Jul/31/qwen3-coder-flash/
- https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct/blob/main/LICENSE
- https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct
- https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct/resolve/main/config.json
- https://huggingface.co/api/models/Qwen/Qwen3-Coder-30B-A3B-Instruct?expand[]=safetensors
- https://ollama.com/library/qwen3-coder/tags