RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?OllamaStart hereAPI pricesVRAM calculatorNew

Can Kimi Linear 48B A3B Instruct run on the RTX 4060 Ti 16GB?

Not fully in VRAM. Even at 2-bit with 8K context, Kimi Linear 48B A3B Instruct needs about 17.3 GB, more than the 15.2 GB usable on the RTX 4060 Ti 16GB. It can still run with offload (slower): Q4_K_M with part of the model in system RAM, about 34–55 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 12 of 26 MoE layers on the RTX 4060 Ti 16GB (about 13.8 GB), and the experts of the other 14 layers in system RAM (about 14.5 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 34–55 tokens/sec (usable for chat), versus a bandwidth ceiling of ~157 tok/s if it fit fully in VRAM.

llama-server -hf bartowski/moonshotai_Kimi-Linear-48B-A3B-Instruct-GGUF:Q4_K_M -c 8192 -ngl 999 --n-cpu-moe 15

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

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •17.2 GB No17.3 GB No18.0 GB No20.8 GB No—
3-bit (Q3_K_M) •22.0 GB No22.1 GB No22.8 GB No25.6 GB No—
4-bit (Q4_K_M) •28.8 GB No28.9 GB No29.7 GB No32.5 GB No—
5-bit (Q5_K_M) •33.5 GB No33.6 GB No34.3 GB No37.2 GB No—
6-bit (Q6_K) •38.5 GB No38.6 GB No39.3 GB No42.2 GB No—
8-bit (Q8_0) •49.5 GB No49.6 GB No50.3 GB No53.2 GB No—
16-bit (F16/BF16) •92.4 GB No92.5 GB No93.2 GB No96.0 GB No—

• real GGUF file size from bartowski/moonshotai_Kimi-Linear-48B-A3B-Instruct-GGUF. Speed is a bandwidth ceiling (288 GB/s), not a benchmark. Methodology.

Hardware that does fit it

RX 7900 XTRTX 3090RTX 3090 TiRTX 4090RX 7900 XTXRTX 5090

What fits instead

"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

Kimi Linear 48B A3B Instruct on every GPU and Mac · Every model that runs on the RTX 4060 Ti 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 (moonshotai/Kimi-Linear-48B-A3B-Instruct), NVIDIA specs. Model data last checked 2026-11-18; refreshed hourly.