Qwen3 30B-A3B (MoE)
Alibaba Qwen · 30.5B parameters (MoE, 3.3B active per token) · 32,768 token max context · Model card
Memory needed
| Quant | Weights | Total @ 8K ctx |
|---|---|---|
| Q3_K_M | 14.9 GB | 16.9 GB |
| Q4_K_M | 18.5 GB | 20.7 GB |
| Q5_K_M | 21.7 GB | 24.0 GB |
| Q6_K | 25.2 GB | 27.5 GB |
| Q8_0 | 32.4 GB | 35.0 GB |
| F16 | 61.0 GB | 64.4 GB |
Approximate GGUF sizes from bits-per-weight; totals include fp16 KV cache and ~0.8 GB overhead.
Can my hardware run it?
| Device | Verdict | Best quant | Est. tok/s (Q4) |
|---|---|---|---|
| NVIDIA GeForce RTX 5090 | Yes | Q6_K | ~272 |
| NVIDIA GeForce RTX 4090 | Yes | Q4_K_M | ~175 |
| AMD Radeon RX 7900 XTX | Yes | Q4_K_M | ~173 |
| NVIDIA GeForce RTX 3090 Ti | Yes | Q4_K_M | ~161 |
| NVIDIA GeForce RTX 3090 | Yes | Q4_K_M | ~148 |
| Apple M5 Max (40-core GPU, 128GB) | Yes | F16 | ~113 |
| Apple M2 Ultra (76-core GPU, 192GB) | Yes | F16 | ~88.5 |
| Apple M3 Ultra (80-core GPU, 512GB) | Yes | F16 | ~86.5 |
| Apple M4 Max (40-core GPU, 128GB) | Yes | F16 | ~77.9 |
| Apple M5 Pro (20-core GPU, 64GB) | Yes | Q8_0 | ~62.2 |
| Apple M3 Max (40-core GPU, 64GB) | Yes | Q8_0 | ~62.2 |
| Apple M1 Max (32-core GPU, 64GB) | Yes | Q8_0 | ~57.4 |
| NVIDIA DGX Spark | Yes | F16 | ~53.7 |
| Apple M4 Pro (20-core GPU, 48GB) | Yes | Q8_0 | ~47.6 |
| Apple M4 (10-core GPU, 24GB) | Yes | Q3_K_M | ~27.1 |
| NVIDIA GeForce RTX 5080 | No | — | — |
| NVIDIA GeForce RTX 5070 Ti | No | — | — |
| NVIDIA GeForce RTX 5070 | No | — | — |
| NVIDIA GeForce RTX 5060 Ti 16GB | No | — | — |
| NVIDIA GeForce RTX 4080 SUPER | No | — | — |
| NVIDIA GeForce RTX 4070 Ti SUPER | No | — | — |
| NVIDIA GeForce RTX 4060 Ti 8GB | No | — | — |
| NVIDIA GeForce RTX 3060 12GB | No | — | — |
| NVIDIA GeForce RTX 3080 10GB | No | — | — |
Speeds estimated from memory bandwidth (calibrated per device with a measured llama.cpp benchmark where available).