Can Phi 3.5 mini instruct run on the 2x RTX 3090?
Yes. At 8K context the best fit is F16, using about 10.6 GB of the 45.6 GB usable on the 2x RTX 3090. Estimated speed ceiling: ~122 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 3.3 GB Fits | 4.8 GB Fits | 13.8 GB Fits | 49.8 GB No | 661 |
| 3-bit (Q3_K_M) • | 3.8 GB Fits | 5.3 GB Fits | 14.3 GB Fits | 50.3 GB No | 479 |
| 4-bit (Q4_K_M) • | 4.2 GB Fits | 5.7 GB Fits | 14.7 GB Fits | 50.7 GB No | 391 |
| 5-bit (Q5_K_M) • | 4.6 GB Fits | 6.1 GB Fits | 15.1 GB Fits | 51.1 GB No | 332 |
| 6-bit (Q6_K) • | 4.9 GB Fits | 6.4 GB Fits | 15.4 GB Fits | 51.4 GB No | 298 |
| 8-bit (Q8_0) • | 5.7 GB Fits | 7.2 GB Fits | 16.2 GB Fits | 52.2 GB No | 230 |
| 16-bit (F16/BF16) | 9.1 GB Fits | 10.6 GB Fits | 19.6 GB Fits | 55.6 GB No | 122 |
• real GGUF file size from bartowski/Phi-3.5-mini-instruct-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.
More
Phi 3.5 mini instruct on every GPU and Mac · Every model that runs on the 2x RTX 3090 · Try other settings in the calculator
Data: Hugging Face Hub (microsoft/Phi-3.5-mini-instruct), NVIDIA specs. Data as of 2026-11-17.