Can I run Qwen3.5 9B locally?
Qwen3.5 9B by Qwen has 9.7B parameters. At 4-bit with an 8K context it needs about 6.4 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track. See all models that fit in 8 GB of VRAM.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Qwen3.5-9B-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 4.4 GB | 4.5 GB | 5.2 GB | 8.2 GB |
| 3-bit (Q3_K_M) • | usable | 5.4 GB | 5.5 GB | 6.2 GB | 9.2 GB |
| 4-bit (Q4_K_M) • | good, most popular | 6.3 GB | 6.4 GB | 7.2 GB | 10.2 GB |
| 5-bit (Q5_K_M) • | very good | 7.1 GB | 7.2 GB | 8.0 GB | 11.0 GB |
| 6-bit (Q6_K) • | near lossless | 7.9 GB | 8.1 GB | 8.8 GB | 11.8 GB |
| 8-bit (Q8_0) • | effectively lossless | 9.9 GB | 10.0 GB | 10.7 GB | 13.7 GB |
| 16-bit (F16/BF16) • | original weights | 17.7 GB | 17.8 GB | 18.6 GB | 21.6 GB |
Which GPUs and Macs run Qwen3.5 9B?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
| Hardware | Memory | Fits? | Best quant | Needs | Max tok/s (est.) | |
|---|---|---|---|---|---|---|
| RTX 3060 12GB | 12 GB | Fits | Q8_0 | 10.0 GB | 38 | details |
| RTX 4060 Ti 16GB | 16 GB | Fits | Q8_0 | 10.0 GB | 30 | details |
| RTX 3090 | 24 GB | Fits | F16 | 17.8 GB | 52 | details |
| RTX 4090 | 24 GB | Fits | F16 | 17.8 GB | 56 | details |
| RTX 5090 | 32 GB | Fits | F16 | 17.8 GB | 100 | details |
| RTX 5070 Ti | 16 GB | Fits | Q8_0 | 10.0 GB | 94 | details |
| RX 7900 XTX | 24 GB | Fits | F16 | 17.8 GB | 54 | details |
| Mac M4 Pro 48GB | 48 GB | Fits | F16 | 17.8 GB | 15 | details |
| Mac M4 Max 128GB | 128 GB | Fits | F16 | 17.8 GB | 30 | details |
| Mac M3 Ultra 512GB | 512 GB | Fits | F16 | 17.8 GB | 46 | details |
| DGX Spark | 128 GB | Fits | F16 | 17.8 GB | 15 | details |
| RTX 4070 | 12 GB | Fits | Q8_0 | 10.0 GB | 53 | details |
| RTX 5060 Ti 16GB | 16 GB | Fits | Q8_0 | 10.0 GB | 47 | details |
| RTX 4060 | 8 GB | Fits | Q4_K_M | 6.4 GB | 48 | details |
| Arc B580 | 12 GB | Fits | Q8_0 | 10.0 GB | 48 | details |
| RTX 3060 Ti | 8 GB | Fits | Q4_K_M | 6.4 GB | 79 | details |
| RTX 3070 | 8 GB | Fits | Q4_K_M | 6.4 GB | 79 | details |
| RTX 3080 10GB | 10 GB | Fits | Q6_K | 8.1 GB | 102 | details |
| RTX 3080 12GB | 12 GB | Fits | Q8_0 | 10.0 GB | 96 | details |
| RTX 3090 Ti | 24 GB | Fits | F16 | 17.8 GB | 56 | details |
| RTX 4060 Ti 8GB | 8 GB | Fits | Q4_K_M | 6.4 GB | 51 | details |
| RTX 4070 Super | 12 GB | Fits | Q8_0 | 10.0 GB | 53 | details |
| RTX 4070 Ti | 12 GB | Fits | Q8_0 | 10.0 GB | 53 | details |
| RTX 4070 Ti Super | 16 GB | Fits | Q8_0 | 10.0 GB | 71 | details |
| RTX 4080 | 16 GB | Fits | Q8_0 | 10.0 GB | 75 | details |
| RTX 4080 Super | 16 GB | Fits | Q8_0 | 10.0 GB | 77 | details |
| RTX 5060 | 8 GB | Fits | Q4_K_M | 6.4 GB | 79 | details |
| RTX 5070 | 12 GB | Fits | Q8_0 | 10.0 GB | 71 | details |
| RTX 5080 | 16 GB | Fits | Q8_0 | 10.0 GB | 101 | details |
| 2x RTX 3090 | 2×24 GB | Fits | F16 | 17.8 GB | 52 | details |
| 2x RTX 4090 | 2×24 GB | Fits | F16 | 17.8 GB | 56 | details |
| 2x RTX 5090 | 2×32 GB | Fits | F16 | 17.8 GB | 100 | details |
| RTX A6000 | 48 GB | Fits | F16 | 17.8 GB | 43 | details |
| RTX 6000 Ada | 48 GB | Fits | F16 | 17.8 GB | 54 | details |
| RTX PRO 6000 Blackwell | 96 GB | Fits | F16 | 17.8 GB | 100 | details |
| L40S | 48 GB | Fits | F16 | 17.8 GB | 48 | details |
| A100 40GB | 40 GB | Fits | F16 | 17.8 GB | 87 | details |
| A100 80GB | 80 GB | Fits | F16 | 17.8 GB | 114 | details |
| H100 PCIe | 80 GB | Fits | F16 | 17.8 GB | 112 | details |
| H100 SXM | 80 GB | Fits | F16 | 17.8 GB | 187 | details |
| H200 | 141 GB | Fits | F16 | 17.8 GB | 268 | details |
| RX 7600 | 8 GB | Fits | Q4_K_M | 6.4 GB | 51 | details |
| RX 7700 XT | 12 GB | Fits | Q8_0 | 10.0 GB | 45 | details |
| RX 7800 XT | 16 GB | Fits | Q8_0 | 10.0 GB | 65 | details |
| RX 7900 GRE | 16 GB | Fits | Q8_0 | 10.0 GB | 60 | details |
| RX 7900 XT | 20 GB | Fits | Q8_0 | 10.0 GB | 84 | details |
| RX 9070 | 16 GB | Fits | Q8_0 | 10.0 GB | 67 | details |
| RX 9070 XT | 16 GB | Fits | Q8_0 | 10.0 GB | 67 | details |
| Ryzen AI Max+ 395 128GB | 128 GB | Fits | F16 | 17.8 GB | 14 | details |
| Arc A770 16GB | 16 GB | Fits | Q8_0 | 10.0 GB | 59 | details |
| Mac M1 16GB | 16 GB | Fits | Q6_K | 8.1 GB | 9.1 | details |
| Mac M2 24GB | 24 GB | Fits | Q8_0 | 10.0 GB | 10 | details |
| Mac M3 24GB | 24 GB | Fits | Q8_0 | 10.0 GB | 10 | details |
| Mac M4 16GB | 16 GB | Fits | Q6_K | 8.1 GB | 16 | details |
| Mac M4 32GB | 32 GB | Fits | F16 | 17.8 GB | 6.7 | details |
| Mac M1 Pro 32GB | 32 GB | Fits | F16 | 17.8 GB | 11 | details |
| Mac M3 Pro 36GB | 36 GB | Fits | F16 | 17.8 GB | 8.4 | details |
| Mac M4 Pro 24GB | 24 GB | Fits | Q8_0 | 10.0 GB | 29 | details |
| Mac M4 Pro 64GB | 64 GB | Fits | F16 | 17.8 GB | 15 | details |
| Mac M1 Max 64GB | 64 GB | Fits | F16 | 17.8 GB | 22 | details |
| Mac M2 Max 96GB | 96 GB | Fits | F16 | 17.8 GB | 22 | details |
| Mac M3 Max 128GB | 128 GB | Fits | F16 | 17.8 GB | 22 | details |
| Mac M4 Max 64GB | 64 GB | Fits | F16 | 17.8 GB | 30 | details |
| Mac M2 Ultra 192GB | 192 GB | Fits | F16 | 17.8 GB | 45 | details |
| Mac M3 Ultra 256GB | 256 GB | Fits | F16 | 17.8 GB | 46 | details |
Prefer an API?
Qwen: Qwen3.5-9B is on OpenRouter at $0.073 per 1M input tokens and $0.107 per 1M output tokens. Price details and history →
Questions
How much VRAM does Qwen3.5 9B need?
About 6.4 GB at 4-bit (Q4_K_M) with an 8K context: 5.3 GB of weights (real GGUF file size), 0.3 GB of KV cache and 0.9 GB of runtime overhead. At 8-bit it needs about 10.0 GB.
What is the smallest GPU that runs Qwen3.5 9B?
At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 (8 GB).
Can I run Qwen3.5 9B on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
Qwen3.5 9B Base
9.7B · 6.4 GB at 4-bitQwen3 VL 8B Instruct
8.8B · 6.5 GB at 4-bitQwen2.5 VL 7B Instruct
8.3B · 5.5 GB at 4-bitOrnith 1.5 9B
9.7B · 6.6 GB at 4-bitSenseNova Loop 9B Preview
9.7B · 6.4 GB at 4-bitMiMo V2.6 Distill Qwen 9B
9.4B · 6.6 GB at 4-bitModel facts
| Hugging Face repo | Qwen/Qwen3.5-9B |
| Parameters | 9,653,104,368 (9.7B) |
| Layers / hidden size | 32 / 4096 |
| Attention | GQA (hybrid: 24 linear/state layers); 16 heads, 4 KV heads, head dim 256 |
| KV cache per token (FP16) | 32.0 KiB |
| Max context | 262,144 tokens |
| License | apache-2.0 |
| Released on HF | Feb 27, 2026 |
| Downloads (30d) / likes | 6,080,782 / 2,384 |
Sources: Hugging Face model API · config.json (Qwen/Qwen3.5-9B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.