Kimi K3 DSpark — Hardware Requirements & GPU Compatibility
Chat
Kimi K3 DSpark is a 2.2B-parameter open language model from RadixArk in the Kimi K3 family. It supports a context window of up to 1,048,576 tokens. At Q4_K_M it needs about 1.72 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 128K)Best on 24 GB (Q4_K_M · ~381 tok/s · 512K)Macs from 8 GB (Q4_K_M · ~42 tok/s)
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Kimi K3 DSpark
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/Anbeeld/Kimi-K3-DSpark-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 21.8 GB free · 16K context
Download: Kimi-K3-DSpark-Q4_K_M.gguf1.33 GBAnbeeld
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 24 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 3080 Ti (12 GB) (9 GB headroom) · Q4_K_M
- Generation speed
- ~345 tok/s
- generation speed
- Cost per 1M output tokens
- $0.098
- per 1M output tokens
How Much VRAM Does Kimi K3 DSpark Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 1.3 GB | 2.4 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 512K context · ~493 tok/s | |
| 1.5 GB | 2.6 GB | 3-bit medium — 3-bit medium quantization | Fits: 512K context · ~446 tok/s | |
| 1.7 GB | 2.8 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 512K context · ~381 tok/s | |
| 2.0 GB | 3.1 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 512K context · ~331 tok/s | |
| 2.2 GB | 3.3 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 512K context · ~294 tok/s | |
| 2.6 GB | 3.7 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 512K context · ~250 tok/s | |
| 4.9 GB | 6 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 512K context · ~135 tok/s |
Selected Q4_K_M
4-bit medium quantization2.2GBat 16K context
- Weights
- 1.4 GB
- KV cache
- 0.6 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 21.8 GB free · ~381 tok/s
≥50 tok/s needs ≥132 GB/s · RTX 5090 ~677, RTX 3090 Ti ~381
Which Hardware Can Run Kimi K3 DSpark?
Q4_K_M · 1.7 GBKimi K3 DSpark (Q4_K_M) requires 1.7 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 3+ GB is recommended. Using the full 1049K context window can add up to 37.5 GB, bringing total usage to 39.2 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Kimi K3 DSpark, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 5.8 GB free · up to 128K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~287Simulate output on RTX 3080RTX 2080 Ti~233Simulate output on RTX 2080 TiRTX 3070 Ti~230Simulate output on RTX 3070 TiM3*8 GB~42Simulate output on iMac 24" M3 (8 GB)Orin Nano~39Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~28Simulate output on iPhone 1720 GB
fits · 17.8 GB free · up to 256K ctx
1 GPU
RX 7900 XT~279Simulate output on RX 7900 XT32–40 GB
fits · 29.8 GB free · up to 512K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~677Simulate output on RTX 5090M5 Max36 GB~187Simulate output on Mac Studio M5 Max (36 GB)M532 GB~63Simulate output on MacBook Air 13" M5 (32 GB)Orin~77Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~51Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Kimi K3 DSpark
Kimi K3 family · 3 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| LFM3 Omni 2BLiquid AI | — | 2.4 GB | ~385 | same size |
| Qwen3.5 2BAlibaba | — | 1.9 GB | ~370 | same size |
| Qwen3 VL 2B InstructAlibaba | — | 3.5 GB | ~362 | same size |
| SmolVLM2 2.2B InstructHugging Face | — | 2.3 GB | ~443 | same size |
| SmolVLM InstructHugging Face | — | 4.9 GB | ~320 | same size |
| Qwen2 VL 2B InstructAlibaba | — | 2.1 GB | ~390 | same size |
Specifications
- Publisher
- RadixArk
- Family
- Kimi K3
- Parameters
- 2.2B
- Architecture
- DSparkDraftModel
- Context Length
- 1,048,576 tokens
- Vocabulary Size
- 163,840
- Release Date
- 2026-07-27
Frequently Asked Questions
How much VRAM does Kimi K3 DSpark need?
Kimi K3 DSpark requires 1.7 GB of VRAM at Q4_K_M, or 4.9 GB at BF16. Full 1049K context adds up to 37.5 GB (39.2 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 2.2B × 4.8 bits ÷ 8 = 1.3 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 37.9 GB (at full 1049K context)
VRAM usage by quantization
What's the best quantization for Kimi K3 DSpark?
For Kimi K3 DSpark, Q4_K_M (1.7 GB) offers the best balance of quality and VRAM usage. Q5_K_M (2.0 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 1.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Kimi K3 DSpark on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run Kimi K3 DSpark. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 1.7 GB of usable unified memory (RAM minus macOS overhead).
Can I run Kimi K3 DSpark locally?
Yes — Kimi K3 DSpark can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Kimi K3 DSpark?
At Q4_K_M, Kimi K3 DSpark can reach ~2791 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~381 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 1.72 × 0.65 = ~3023 tok/s
Estimated speed at Q4_K_M (1.7 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Kimi K3 DSpark?
At Q4_K_M, the download is about 1.35 GB. The full-precision BF16 version is 4.50 GB. The smallest option (Q2_K) is 0.96 GB.
Which GPUs can run Kimi K3 DSpark?
52 consumer GPUs can run Kimi K3 DSpark at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Kimi K3 DSpark?
163 devices with unified memory can run Kimi K3 DSpark at Q4_K_M (1.7 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.