Can Ling 3.0 Flash Sante run on the RX 9070?
Not fully in VRAM. Even at 2-bit with 8K context, Ling 3.0 Flash Sante needs about 47.2 GB, more than the 15.2 GB usable on the RX 9070. It can still run with offload (slower): Q4_K_M with part of the model in system RAM, about 20–37 tokens/sec (estimate). Details and command.
Runs with offload (slower)
Runs with offload 4-bit (Q4_K_M) at 8K context: all attention layers plus the experts of 6 of 41 MoE layers on the RX 9070 (about 13.2 GB), and the experts of the other 35 layers in system RAM (about 62.1 GB). Needs a PC with at least 128 GB of system RAM. Estimated speed: 20–37 tokens/sec (usable for chat), versus a bandwidth ceiling of ~181 tok/s if it fit fully in VRAM.
llama-server -hf bartowski/Ling-3.0-flash-Sante-GGUF:Q4_K_M -c 8192 -ngl 999 --n-cpu-moe 36
Estimate, not a benchmark. The command downloads the GGUF from Hugging Face. The speed range spans dual-channel DDR4-3200 to DDR5-6000 system RAM; your CPU, RAM speed and llama.cpp version move it. If llama.cpp runs out of GPU memory, raise --n-cpu-moe by 1-2; if VRAM is left over, lower it for more speed. Recent llama.cpp builds can also pick the split automatically. How we estimate offload.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 47.0 GB No | 47.2 GB No | 48.3 GB No | 52.7 GB No | — |
| 3-bit (Q3_K_M) • | 58.2 GB No | 58.3 GB No | 59.5 GB No | 63.9 GB No | — |
| 4-bit (Q4_K_M) • | 75.8 GB No | 75.9 GB No | 77.0 GB No | 81.5 GB No | — |
| 5-bit (Q5_K_M) • | 91.8 GB No | 92.0 GB No | 93.1 GB No | 97.5 GB No | — |
| 6-bit (Q6_K) • | 105 GB No | 105 GB No | 106 GB No | 110 GB No | — |
| 8-bit (Q8_0) • | 127 GB No | 127 GB No | 128 GB No | 133 GB No | — |
| 16-bit (F16/BF16) • | 238 GB No | 239 GB No | 240 GB No | 244 GB No | — |
• real GGUF file size from bartowski/Ling-3.0-flash-Sante-GGUF. Speed is a bandwidth ceiling (640 GB/s), not a benchmark. Methodology.
Hardware that does fit it
Mac M4 Pro 64GBMac M1 Max 64GBMac M4 Max 64GBA100 80GBH100 PCIeH100 SXM
What fits instead
- Cheapest listed GPU that runs Ling 3.0 Flash Sante: DGX Spark (128 GB, launch MSRP $3,999) at Q5_K_M, fully in VRAM at 8K context.
- Largest popular model that fits the RX 9070: gpt-oss 20B (21B, MoE) at Q8_0, about 12.3 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
Ling 3.0 Flash Sante on every GPU and Mac · Every model that runs on the RX 9070 · Other models that fit 16 GB VRAM · Try other settings in the calculator
Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs
Data: Hugging Face Hub (inclusionAI/Ling-3.0-flash-Sante), AMD specs. Model data last checked 2026-11-18; refreshed hourly.