Llama 3.2 3B hardware requirements
3.21B parameters. Needs about 3.5 GB at Q4_K_M with 8K context. Model card.
Memory needed at each quantization
| Quant | Quality | 8K context | 32K context |
|---|---|---|---|
| FP16 | Full quality | 8 GB | 10.8 GB |
| Q8_0 | Near-lossless | 5 GB | 7.8 GB |
| Q6_K | Excellent | 4.2 GB | 7 GB |
| Q5_K_M | Very good | 3.8 GB | 6.6 GB |
| Q4_K_M | Good, most popular | 3.5 GB | 6.3 GB |
| Q3_K_M | Noticeably degraded | 3.1 GB | 5.9 GB |
| Q2_K | Heavily degraded | 2.7 GB | 5.6 GB |
Run it with Ollama: ollama run llama3.2:3b (downloads Q4_K_M by default).
Run it in the cloud
Don't want to buy hardware? Run Llama 3.2 3B on a rented GPU today.
Rent a cloud GPU by the hour. No upfront cost, cancel any time.
Every GPU and Mac checked
| GPU | Runs? | Best fit |
|---|---|---|
| NVIDIA RTX 3060 12GB | Yes | FP16 · ~8 GB · ~34 tok/s |
| NVIDIA RTX 3070 | Yes | FP16 · ~8 GB · ~42 tok/s |
| NVIDIA RTX 3080 10GB | Yes | FP16 · ~8 GB · ~71 tok/s |
| NVIDIA RTX 3090 | Yes | FP16 · ~8 GB · ~87 tok/s |
| NVIDIA RTX 4060 | Yes | FP16 · ~8 GB · ~25 tok/s |
| NVIDIA RTX 4060 Ti 16GB | Yes | FP16 · ~8 GB · ~27 tok/s |
| NVIDIA RTX 5060 Ti 16GB | Yes | FP16 · ~8 GB · ~42 tok/s |
| NVIDIA RTX 4070 | Yes | FP16 · ~8 GB · ~47 tok/s |
| NVIDIA RTX 4070 Ti Super | Yes | FP16 · ~8 GB · ~63 tok/s |
| NVIDIA RTX 4080 | Yes | FP16 · ~8 GB · ~67 tok/s |
| NVIDIA RTX 4090 | Yes | FP16 · ~8 GB · ~94 tok/s |
| NVIDIA RTX 5070 | Yes | FP16 · ~8 GB · ~63 tok/s |
| NVIDIA RTX 5070 Ti | Yes | FP16 · ~8 GB · ~84 tok/s |
| NVIDIA RTX 5080 | Yes | FP16 · ~8 GB · ~90 tok/s |
| NVIDIA RTX 5090 | Yes | FP16 · ~8 GB · ~167 tok/s |
| AMD Radeon RX 7800 XT | Yes | FP16 · ~8 GB · ~58 tok/s |
| AMD Radeon RX 7900 XTX | Yes | FP16 · ~8 GB · ~90 tok/s |
| Apple M4 (16GB) | Yes | FP16 · ~8 GB · ~11 tok/s |
| Apple M1 (16GB) | Yes | FP16 · ~8 GB · ~6 tok/s |
| Apple M2 Pro (32GB) | Yes | FP16 · ~8 GB · ~19 tok/s |
| Apple M3 Max (64GB) | Yes | FP16 · ~8 GB · ~37 tok/s |
| Apple M4 Pro (48GB) | Yes | FP16 · ~8 GB · ~26 tok/s |
| Apple M4 Max (128GB) | Yes | FP16 · ~8 GB · ~51 tok/s |
| Apple M2 Ultra (192GB) | Yes | FP16 · ~8 GB · ~75 tok/s |