Can phi 2 run on the A100 40GB?
Yes. At 8K context the best fit is F16, using about 8.2 GB of the 38.0 GB usable on the A100 40GB. Estimated speed ceiling: ~280 tokens/sec.
Memory by quant and context
| Quant | Max tok/s (est., 8K) |
|---|---|
| 2-bit (Q2_K) • | 1325 |
| 3-bit (Q3_K_M) • | 1051 |
| 4-bit (Q4_K_M) • | 869 |
| 5-bit (Q5_K_M) • | 750 |
| 6-bit (Q6_K) • | 681 |
| 8-bit (Q8_0) • | 526 |
| 16-bit (F16/BF16) | 280 |
• real GGUF file size from TheBloke/phi-2-GGUF. Speed is a bandwidth ceiling (1555 GB/s), not a benchmark. Methodology.
More
phi 2 on every GPU and Mac · Every model that runs on the A100 40GB · Other models that fit 48 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (microsoft/phi-2), NVIDIA specs. Data as of 2026-11-17.