Why this recommendation
- Minimum GPU VRAM Capacity
- Usable VRAM Target
- Usable VRAM on Recommended Hardware Tier
- Fit Headroom
- Estimated Speed
Values Used In Calculations
Formula used
Assumptions used
Hardware tier
single accelerator
≤ 24 Common consumer / small inference
- 8 GB RTX 4060
- 12 GB RTX 3060 RTX 4070 RX 9070 GRE RX 6700 XT
- 16 GB RTX 4080 RX 7800 XT RTX 5000 Ada Mac mini M4 16 GB Cloud TPU v5e
- 20 GB RX 7900 XT
- 24 GB RTX 4090 RTX 3090 L4 Mac mini M4 24 GB
32–48 High-end consumer, workstation, larger local inference
64–96 Large local systems and datacenter accelerators
- 64 GB Mac Studio M4 Max 64 GB AMD Instinct MI210
- 80 GB A100 H100
- 95 GB Cloud TPU v5p
- 96 GB Mac Studio M3 Ultra 96 GB RTX PRO 6000 Blackwell
141–192 High-memory datacenter / frontier accelerators
- 141 GB H200
- 180 GB B200
- 192 GB Cloud TPU7x
> 192 Does not fit on a single accelerator
- 160 GB 2x 80 GB GPUs with tensor/model parallelism
- 320 GB 4x 80 GB GPUs with tensor/model parallelism