Can deepseek moe 16b base run on the RX 7900 GRE?
Yes. At 8K context the best fit is Q4_K_M, using about 12.5 GB of the 15.2 GB usable on the RX 7900 GRE. Estimated speed ceiling: ~307 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | Max tok/s (est., 8K) |
|---|---|---|
| 2-bit (Q2_K) • | 7.7 GB Fits | 497 |
| 3-bit (Q3_K_M) • | 9.4 GB Fits | 393 |
| 4-bit (Q4_K_M) • | 11.6 GB Fits | 307 |
| 5-bit (Q5_K_M) • | 13.0 GB Fits | 269 |
| 6-bit (Q6_K) • | 15.1 GB Tight | — |
| 8-bit (Q8_0) • | 17.7 GB No | — |
| 16-bit (F16/BF16) | 32.0 GB No | — |
• real GGUF file size from mradermacher/deepseek-moe-16b-base-GGUF. Speed is a bandwidth ceiling (576 GB/s), not a benchmark. Methodology.
More
deepseek moe 16b base on every GPU and Mac · Every model that runs on the RX 7900 GRE · Other models that fit 16 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (deepseek-ai/deepseek-moe-16b-base), AMD specs. Data as of 2026-11-17.