Can DeepSeek V4 Flash 0731 run on the Arc A770 16GB?
Not fully in VRAM. Even at 2-bit with 8K context, DeepSeek V4 Flash 0731 needs about 111 GB, more than the 15.2 GB usable on the Arc A770 16GB. It can still run with offload (slower): Q2_K with part of the model in system RAM, about 7.9–14 tokens/sec (estimate). Details and command.
Runs with offload (slower)
Runs with offload 2-bit (Q2_K) at 8K context: all attention layers plus the experts of 2 of 44 MoE layers on the Arc A770 16GB (about 12.8 GB), and the experts of the other 42 layers in system RAM (about 97.2 GB). Needs a PC with at least 128 GB of system RAM. Estimated speed: 7.9–14 tokens/sec (slow but usable), versus a bandwidth ceiling of ~53 tok/s if it fit fully in VRAM.
llama-server -hf ggml-org/DeepSeek-V4-Flash-0731-GGUF:Q2_K -c 8192 -ngl 999 --n-cpu-moe 42
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.
Newer compressed-attention design: long-context figures are an upper bound.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 110 GB No | 111 GB No | 113 GB No | 122 GB No | — |
| 3-bit (Q3_K_M) | 139 GB No | 140 GB No | 142 GB No | 151 GB No | — |
| 4-bit (Q4_K_M) | 174 GB No | 175 GB No | 177 GB No | 186 GB No | — |
| 5-bit (Q5_K_M) | 203 GB No | 203 GB No | 206 GB No | 215 GB No | — |
| 6-bit (Q6_K) | 233 GB No | 234 GB No | 236 GB No | 245 GB No | — |
| 8-bit (Q8_0) | 302 GB No | 302 GB No | 305 GB No | 314 GB No | — |
| 16-bit (F16/BF16) | 568 GB No | 568 GB No | 570 GB No | 579 GB No | — |
• real GGUF file size from ggml-org/DeepSeek-V4-Flash-0731-GGUF. Speed is a bandwidth ceiling (560 GB/s), not a benchmark. Methodology.
Hardware that does fit it
H200Mac M2 Ultra 192GBMac M3 Ultra 256GBMac M3 Ultra 512GB
What fits instead
- No listed GPU with a list price runs DeepSeek V4 Flash 0731 fully in memory. The smallest listed hardware that does is the H200 (141 GB).
- Largest popular model that fits the Arc A770 16GB: 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
DeepSeek V4 Flash 0731 on every GPU and Mac · Every model that runs on the Arc A770 16GB · 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 (deepseek-ai/DeepSeek-V4-Flash-0731), Intel specs. Model data last checked 2026-11-18; refreshed hourly.