✅ — TPU v6e Can Run Llama 3.2 1B
The TPU v6e packs 256GB VRAM. Llama 3.2 1B needs ~0.6GB. You're good to go.
✅ Plenty of headroom
Hardware Specs
- GPU
- TPU v6e
- VRAM
- 256 GB
- Bandwidth
- 4500 GB/s
- MSRP
- $0
Model Requirements
- Model
- Llama 3.2 1B
- Parameters
- 1B
- VRAM (Q4)
- ~0.6 GB
- Context
- N/A
Estimated Performance
6750 tok/s
Estimated inference speed (Q4 quantization)