Can DeepSeek V4 Pro 0813 run on the RTX 4060 Ti 8GB?
No. Even at 2-bit with 8K context, DeepSeek V4 Pro 0813 needs about 549 GB, more than the 7.6 GB usable on the RTX 4060 Ti 8GB.
Even with llama.cpp CPU offload (-ngl, --n-cpu-moe) and 128 GB of system RAM, this model does not fit at 2-bit, or the card cannot hold a useful share of it. See the offload method.
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) • | 548 GB No | 549 GB No | 552 GB No | 565 GB No | — |
| 3-bit (Q3_K_M) • | 712 GB No | 713 GB No | 716 GB No | 729 GB No | — |
| 4-bit (Q4_K_M) | 941 GB No | 941 GB No | 945 GB No | 957 GB No | — |
| 5-bit (Q5_K_M) | 1096 GB No | 1097 GB No | 1100 GB No | 1113 GB No | — |
| 6-bit (Q6_K) | 1262 GB No | 1262 GB No | 1265 GB No | 1278 GB No | — |
| 8-bit (Q8_0) | 1634 GB No | 1635 GB No | 1638 GB No | 1651 GB No | — |
| 16-bit (F16/BF16) | 3075 GB No | 3076 GB No | 3079 GB No | 3092 GB No | — |
• real GGUF file size from qtum/DeepSeek-V4-Pro-0813-GGUF. Speed is a bandwidth ceiling (288 GB/s), not a benchmark. Methodology.
What fits instead
- Nothing we list runs DeepSeek V4 Pro 0813 fully in memory at 8K context. It needs a multi-GPU server.
- Largest popular model that fits the RTX 4060 Ti 8GB: Qwen3.5 9B (9.7B) at Q4_K_M, about 6.4 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 Pro 0813 on every GPU and Mac · Every model that runs on the RTX 4060 Ti 8GB · Other models that fit 8 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-Pro-0813), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.