Can Phi 3 mini 128k instruct run on the RX 9070 XT?
Yes. At 8K context the best fit is F16, using about 10.6 GB of the 15.2 GB usable on the RX 9070 XT. Estimated speed ceiling: ~84 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 Tight | 49.8 GB No | 452 |
| 3-bit (Q3_K_M) • | 3.8 GB Fits | 5.3 GB Fits | 14.3 GB Tight | 50.3 GB No | 327 |
| 4-bit (Q4_K_M) • | 4.2 GB Fits | 5.7 GB Fits | 14.7 GB Tight | 50.7 GB No | 267 |
| 5-bit (Q5_K_M) • | 4.6 GB Fits | 6.1 GB Fits | 15.1 GB Tight | 51.1 GB No | 227 |
| 6-bit (Q6_K) • | 4.9 GB Fits | 6.4 GB Fits | 15.4 GB No | 51.4 GB No | 204 |
| 8-bit (Q8_0) • | 5.7 GB Fits | 7.2 GB Fits | 16.2 GB No | 52.2 GB No | 158 |
| 16-bit (F16/BF16) | 9.1 GB Fits | 10.6 GB Fits | 19.6 GB No | 55.6 GB No | 84 |
• real GGUF file size from bartowski/Phi-3.1-mini-128k-instruct-GGUF. Speed is a bandwidth ceiling (640 GB/s), not a benchmark. Methodology.
More
Phi 3 mini 128k instruct on every GPU and Mac · Every model that runs on the RX 9070 XT · Other models that fit 16 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (microsoft/Phi-3-mini-128k-instruct), AMD specs. Data as of 2026-11-17.