Can I run Qwen3.5 2B locally?
Qwen3.5 2B by Qwen has 2.3B parameters. At 4-bit with an 8K context it needs about 2.2 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/Qwen3.5-2B-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 | 1.7 GB | 1.8 GB | 2.0 GB | 3.2 GB |
| 3-bit (Q3_K_M) • | usable | 2.0 GB | 2.0 GB | 2.3 GB | 3.4 GB |
| 4-bit (Q4_K_M) • | good, most popular | 2.1 GB | 2.2 GB | 2.4 GB | 3.6 GB |
| 5-bit (Q5_K_M) • | very good | 2.3 GB | 2.3 GB | 2.6 GB | 3.7 GB |
| 6-bit (Q6_K) • | near lossless | 2.4 GB | 2.4 GB | 2.7 GB | 3.8 GB |
| 8-bit (Q8_0) • | effectively lossless | 2.8 GB | 2.8 GB | 3.1 GB | 4.2 GB |
| 16-bit (F16/BF16) • | original weights | 4.4 GB | 4.5 GB | 4.8 GB | 5.9 GB |
Which GPUs and Macs run Qwen3.5 2B?
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 | F16 | 4.5 GB | 95 | details |
| RTX 4060 Ti 16GB | 16 GB | Fits | F16 | 4.5 GB | 76 | details |
| RTX 3090 | 24 GB | Fits | F16 | 4.5 GB | 248 | details |
| RTX 4090 | 24 GB | Fits | F16 | 4.5 GB | 267 | details |
| RTX 5090 | 32 GB | Fits | F16 | 4.5 GB | 475 | details |
| RTX 5070 Ti | 16 GB | Fits | F16 | 4.5 GB | 237 | details |
| RX 7900 XTX | 24 GB | Fits | F16 | 4.5 GB | 254 | details |
| Mac M4 Pro 48GB | 48 GB | Fits | F16 | 4.5 GB | 72 | details |
| Mac M4 Max 128GB | 128 GB | Fits | F16 | 4.5 GB | 145 | details |
| Mac M3 Ultra 512GB | 512 GB | Fits | F16 | 4.5 GB | 217 | details |
| DGX Spark | 128 GB | Fits | F16 | 4.5 GB | 72 | details |
| RTX 4070 | 12 GB | Fits | F16 | 4.5 GB | 133 | details |
| RTX 5060 Ti 16GB | 16 GB | Fits | F16 | 4.5 GB | 119 | details |
| RTX 4060 | 8 GB | Fits | F16 | 4.5 GB | 72 | details |
| Arc B580 | 12 GB | Fits | F16 | 4.5 GB | 121 | details |
| RTX 3060 Ti | 8 GB | Fits | F16 | 4.5 GB | 119 | details |
| RTX 3070 | 8 GB | Fits | F16 | 4.5 GB | 119 | details |
| RTX 3080 10GB | 10 GB | Fits | F16 | 4.5 GB | 201 | details |
| RTX 3080 12GB | 12 GB | Fits | F16 | 4.5 GB | 242 | details |
| RTX 3090 Ti | 24 GB | Fits | F16 | 4.5 GB | 267 | details |
| RTX 4060 Ti 8GB | 8 GB | Fits | F16 | 4.5 GB | 76 | details |
| RTX 4070 Super | 12 GB | Fits | F16 | 4.5 GB | 133 | details |
| RTX 4070 Ti | 12 GB | Fits | F16 | 4.5 GB | 133 | details |
| RTX 4070 Ti Super | 16 GB | Fits | F16 | 4.5 GB | 178 | details |
| RTX 4080 | 16 GB | Fits | F16 | 4.5 GB | 190 | details |
| RTX 4080 Super | 16 GB | Fits | F16 | 4.5 GB | 195 | details |
| RTX 5060 | 8 GB | Fits | F16 | 4.5 GB | 119 | details |
| RTX 5070 | 12 GB | Fits | F16 | 4.5 GB | 178 | details |
| RTX 5080 | 16 GB | Fits | F16 | 4.5 GB | 254 | details |
| 2x RTX 3090 | 2×24 GB | Fits | F16 | 4.5 GB | 248 | details |
| 2x RTX 4090 | 2×24 GB | Fits | F16 | 4.5 GB | 267 | details |
| 2x RTX 5090 | 2×32 GB | Fits | F16 | 4.5 GB | 475 | details |
| RTX A6000 | 48 GB | Fits | F16 | 4.5 GB | 203 | details |
| RTX 6000 Ada | 48 GB | Fits | F16 | 4.5 GB | 254 | details |
| RTX PRO 6000 Blackwell | 96 GB | Fits | F16 | 4.5 GB | 475 | details |
| L40S | 48 GB | Fits | F16 | 4.5 GB | 229 | details |
| A100 40GB | 40 GB | Fits | F16 | 4.5 GB | 412 | details |
| A100 80GB | 80 GB | Fits | F16 | 4.5 GB | 540 | details |
| H100 PCIe | 80 GB | Fits | F16 | 4.5 GB | 530 | details |
| H100 SXM | 80 GB | Fits | F16 | 4.5 GB | 887 | details |
| H200 | 141 GB | Fits | F16 | 4.5 GB | 1271 | details |
| RX 7600 | 8 GB | Fits | F16 | 4.5 GB | 76 | details |
| RX 7700 XT | 12 GB | Fits | F16 | 4.5 GB | 114 | details |
| RX 7800 XT | 16 GB | Fits | F16 | 4.5 GB | 165 | details |
| RX 7900 GRE | 16 GB | Fits | F16 | 4.5 GB | 153 | details |
| RX 7900 XT | 20 GB | Fits | F16 | 4.5 GB | 212 | details |
| RX 9070 | 16 GB | Fits | F16 | 4.5 GB | 170 | details |
| RX 9070 XT | 16 GB | Fits | F16 | 4.5 GB | 170 | details |
| Ryzen AI Max+ 395 128GB | 128 GB | Fits | F16 | 4.5 GB | 68 | details |
| Arc A770 16GB | 16 GB | Fits | F16 | 4.5 GB | 148 | details |
| Mac M1 16GB | 16 GB | Fits | F16 | 4.5 GB | 18 | details |
| Mac M2 24GB | 24 GB | Fits | F16 | 4.5 GB | 26 | details |
| Mac M3 24GB | 24 GB | Fits | F16 | 4.5 GB | 26 | details |
| Mac M4 16GB | 16 GB | Fits | F16 | 4.5 GB | 32 | details |
| Mac M4 32GB | 32 GB | Fits | F16 | 4.5 GB | 32 | details |
| Mac M1 Pro 32GB | 32 GB | Fits | F16 | 4.5 GB | 53 | details |
| Mac M3 Pro 36GB | 36 GB | Fits | F16 | 4.5 GB | 40 | details |
| Mac M4 Pro 24GB | 24 GB | Fits | F16 | 4.5 GB | 72 | details |
| Mac M4 Pro 64GB | 64 GB | Fits | F16 | 4.5 GB | 72 | details |
| Mac M1 Max 64GB | 64 GB | Fits | F16 | 4.5 GB | 106 | details |
| Mac M2 Max 96GB | 96 GB | Fits | F16 | 4.5 GB | 106 | details |
| Mac M3 Max 128GB | 128 GB | Fits | F16 | 4.5 GB | 106 | details |
| Mac M4 Max 64GB | 64 GB | Fits | F16 | 4.5 GB | 145 | details |
| Mac M2 Ultra 192GB | 192 GB | Fits | F16 | 4.5 GB | 212 | details |
| Mac M3 Ultra 256GB | 256 GB | Fits | F16 | 4.5 GB | 217 | details |
Questions
How much VRAM does Qwen3.5 2B need?
About 2.2 GB at 4-bit (Q4_K_M) with an 8K context: 1.2 GB of weights (real GGUF file size), 0.1 GB of KV cache and 0.9 GB of runtime overhead. At 8-bit it needs about 2.8 GB.
What is the smallest GPU that runs Qwen3.5 2B?
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 2B 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 1.7B
2.0B · 2.6 GB at 4-bitQwen2.5 1.5B Instruct
1.5B · 1.9 GB at 4-bitQwen2.5 3B Instruct
3.1B · 2.9 GB at 4-bitSmolVLM Instruct
2.2B · 3.0 GB at 4-bitWeVisOne VL 2B
2.4B · 2.8 GB at 4-bitMiniCPM5 2B
2.5B · 2.5 GB at 4-bitModel facts
| Hugging Face repo | Qwen/Qwen3.5-2B |
| Parameters | 2,274,069,824 (2.3B) |
| Layers / hidden size | 24 / 2048 |
| Attention | GQA (hybrid: 18 linear/state layers); 8 heads, 2 KV heads, head dim 256 |
| KV cache per token (FP16) | 12.0 KiB |
| Max context | 262,144 tokens |
| License | apache-2.0 |
| Released on HF | Feb 28, 2026 |
| Downloads (30d) / likes | 4,915,024 / 487 |
Sources: Hugging Face model API · config.json (Qwen/Qwen3.5-2B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.