text model · Kimi · Windows
Can I run Kimi K2 Instruct on Nvidia GeForce RTX 3090 (24GB)?
No. Kimi K2 Instruct needs ~586.6 GB even at Q4_K_M, but Nvidia GeForce RTX 3090 (24GB) only has ~23 GB usable.
Needs ~586.6 GB even at Q4_K_M, but only ~23 GB is usable.
That figure is at a 4k context and moves about ±15% as context length changes.
The gap is about 563.6 GB: Kimi K2 Instruct needs roughly 586.6 GB at Q4_K_M and Nvidia GeForce RTX 3090 (24GB) leaves only about 23 GB usable for a model. No single tracked device has enough memory; Kimi K2 Instruct needs a multi-GPU or high-memory rig.
- Q4_K_M needed
- ~586.6 GB
- Usable on device
- ~23 GB
- Device memory
- 24 GB
Which quant fits
- Parameters
- 1000B (MoE, 32B active)
- Q4_K_M size
- 578.15 GB
- Q8_0 size
- 1016.12 GB
- Context
- 128k
- Ollama tag
- kimi-k2
- Memory
- 24 GB vram
- Usable for weights
- ~23 GB
- Power draw
- ~350 W
- Best runtime
- vLLM (Linux) / Ollama (CUDA)
What you can run instead
FAQ
Can Nvidia GeForce RTX 3090 (24GB) run Kimi K2 Instruct?
No. Kimi K2 Instruct needs ~586.6 GB even at Q4_K_M, but Nvidia GeForce RTX 3090 (24GB) only has ~23 GB usable.
How much memory does Kimi K2 Instruct need?
Nvidia GeForce RTX 3090 (24GB) does not have enough memory. At Q4_K_M the weights are ~578.15 GB; with KV cache and runtime overhead, budget ~586.6 GB at a 4k context. It is a Mixture-of-Experts model (1000B total / 32B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run Kimi K2 Instruct on Windows?
LM Studio for a simple setup; Ollama (CUDA) for the most speed. AMD GPUs run via Vulkan/ROCm at roughly half CUDA throughput. NVIDIA is the smooth path on Windows.
Embed this
A live badge for your model card or README, updated as the data is. [](https://localmodel.run/can-i-run/kimi-k2/nvidia-rtx-3090-24gb) Sources
Weights are measured from GGUF files; KV cache and overhead are computed, so totals can vary ~15% with context and runtime. Any tok/s is a bandwidth estimate. See methodology.