⌘K
← All AI Models

Can you run Kimi-Dev-72B locally?

Kimi family · 2026 · 72.7B parameters (72.7B activated per token) · Hugging Face model card

Kimi-Dev-72B is a dense open-source coding model for software engineering and issue resolution, continued from Qwen2.5-72B (HF safetensors.total=72,706,203,648). Native context is 131,072 tokens per config max_position_embeddings. Architecture is qwen2 GQA, 80 layers, 8 KV heads, MIT licensed. The vendor card reports 60.4% on SWE-bench Verified; that is a vendor claim and is not stored as measured site data. Local VRAM fit at Q4-class is roughly 48 GB (calculated). Measured speeds: not yet published here.

Minimum: 48 GB+ memory (Q4-class on 72.71B dense, calculated)  ·  Recommended: 64 GB+ for comfortable context headroom (Q4-class on 72.71B dense, calculated)

Parameters (total)
72.7B
Activated per token
72.7B (MoE)
Context window
131,072 tokens
Architecture source
HF config.json (verified)

How much VRAM does Kimi-Dev-72B need at each quantization?

Kimi-Dev-72B needs 48.0 GB of VRAM at Q4_K_M. The table below lists weights-only size and total VRAM including overhead for each common quantization level with 4k and 32k token contexts.

QuantizationBits / weightWeights onlyTotal + KV @4k ctxTotal + KV @32k ctx
Q4_K_M 4.8 43.6 GB 48.0 GB 48.0 GB
Q5_K_M 5.7 51.8 GB 57.0 GB 57.0 GB
Q6_K 6.6 60.0 GB 66.0 GB 66.0 GB
Q8_0 8.5 77.3 GB 85.0 GB 85.0 GB
FP16 16 145.4 GB 160.0 GB 160.0 GB

Method: weights = parameters × bits-per-weight (Q4_K_M ≈ 4.8, Q5_K_M ≈ 5.7, Q6_K ≈ 6.6, Q8_0 ≈ 8.5, FP16 = 16), plus 10% loading overhead, plus KV cache from the verified architecture config (80 layers, 8 KV heads, head dim). Source: HF-verified 2026-09-23 (rechecked 2026-09-27) via api (gated=false private=false params=72.71B safetensors.total=72706203648 downloadable_safetensors_sum=145412520096 ctx=131072 license=mit repo=moonshotai/Kimi-Dev-72B@8791d7981945752a51f692d66f2bbfb3573c9722 layers=80 hidden=8192 kv_heads=8 dtype=bfloat16). dense_or_moe=dense num_experts=None num_experts_per_tok=None active_params_b=72.71 (vendor card; VRAM fit uses active, not dense-equivalent). VRAM figures are documented calculations (Q4 ~0.6GB/B), not measurements. No tok/s invented.

Which GPUs can run Kimi-Dev-72B locally?

At Q4_K_M with a 4k context, Kimi-Dev-72B needs 48.0 GB of VRAM. The lists below are computed live from our GPU database and grouped by how much headroom the card has. Disclosure: CompareAIHardware.com participates in the Amazon Associates program and earns from qualifying purchases through links on this page. Affiliate relationships do not influence our recommendations.

Runs Kimi-Dev-72B comfortably (25%+ VRAM headroom)

GPUVRAMBandwidthType
H100 SXM 80 GB 3350 GB/s Data Center GPU
NVIDIA A100 80GB SXM 80 GB 2039 GB/s Data Center GPU
NVIDIA H100 PCIe 80GB 80 GB 2039 GB/s Data Center GPU
NVIDIA RTX PRO 6000 Blackwell 96 GB 1792 GB/s Pro GPU Check price
H200 SXM 141 GB 4800 GB/s Data Center GPU
AMD Instinct MI300X 192 GB 5300 GB/s Data Center GPU
NVIDIA B200 192 GB 8000 GB/s Data Center GPU
AMD Instinct MI325X 256 GB 6000 GB/s Data Center GPU
AMD Instinct MI355X 288 GB 8000 GB/s Data Center GPU
NVIDIA B300 (Blackwell Ultra) 288 GB 8000 GB/s Data Center GPU

Minimum GPUs that fit Kimi-Dev-72B at Q4_K_M

These GPUs hold the model but leave little headroom — keep contexts short.

GPUVRAMBandwidthType
AMD Radeon Pro W7900 48 GB 864 GB/s Pro GPU Check price
NVIDIA RTX PRO 5000 Blackwell 48 GB 1344 GB/s Pro GPU
RTX 6000 Ada 48 GB 960 GB/s Pro GPU Check price
RTX A6000 48 GB 768 GB/s Pro GPU Check price

Needs 2+ GPUs to run Kimi-Dev-72B

One of these cards is too small on its own, but a pair (tensor or pipeline parallel, ~90% efficiency) covers the 48.0 GB requirement. See our multi-GPU guide for setup.

GPUVRAMBandwidthType
AMD Radeon Pro W7800 32 GB ×2 576 GB/s Pro GPU Check price
GeForce RTX 5090 32 GB ×2 1792 GB/s Consumer GPU Check price
NVIDIA RTX 5000 Ada 32 GB ×2 576 GB/s Pro GPU Check price
NVIDIA RTX PRO 4500 Blackwell 32 GB ×2 896 GB/s Pro GPU Check price
NVIDIA A100 40GB SXMEOL 40 GB ×2 1555 GB/s Data Center GPU

Can a Mac run Kimi-Dev-72B?

Yes — these Apple Silicon machines fit Kimi-Dev-72B at Q4_K_M, since macOS lets the GPU use about 75% of unified memory. Generation speed is bound by memory bandwidth, so the GB/s column matters as much as capacity.

Apple SiliconUnified memoryUsable by GPU (~75%)Bandwidth
Apple M4 Pro 64 GB 48 GB 273 GB/s
M5 Pro (Mac mini) 64 GB 48 GB 307 GB/s
Apple M3 Max 128 GB 96 GB 400 GB/s
M4 Max (MacBook Pro) 128 GB 96 GB 546 GB/s
M5 Max (Mac Studio) 128 GB 96 GB —
M5 Ultra (Mac Studio) 512 GB 384 GB 1200 GB/s

Can Kimi-Dev-72B run on mini PCs, Jetson, or NPU devices?

These edge and NPU devices from our database have enough memory for Kimi-Dev-72B at Q4_K_M. Their memory bandwidth is far below discrete GPUs, so expect a fraction of desktop generation speed.

DeviceMemoryBandwidthType
Jetson AGX Orin 64GB 64 GB 204 GB/s Edge Compute Device
Snapdragon X Elite (X1E-84-100) 64 GB 135 GB/s NPU Chip
AMD Ryzen AI 9 HX 370 (Strix Point) 96 GB 120 GB/s NPU Chip

Which pre-built systems can run Kimi-Dev-72B?

These mini PCs and workstations from our database fit Kimi-Dev-72B at Q4_K_M. Usable-memory figures are conservative: Windows shares about half of system RAM with the GPU by default (Linux can expose more), macOS lets Apple Silicon GPUs use about 75% of unified memory, and Linux unified-memory systems such as GB10 expose roughly 90%. Generation speed is bound by memory bandwidth, so compare the GB/s column before buying.

SystemTypeMemoryUsable for AIBandwidthGPUs
ASRock NUC BOX-255H (Core Ultra 7 255H) Mini PC 96 GB 48 GB 102 GB/s —
Apple Mac Studio M2 Ultra (64GB) Workstation 64 GB 48 GB 800 GB/s 1
Custom Dual RTX 4090 Training Workstation Workstation 48 GB 48 GB 1008 GB/s 2
Dell Precision 7960 Tower (1x RTX 6000 Ada) Workstation 48 GB 48 GB 960 GB/s 1
HP Z8 Fury G5 (1x RTX 6000 Ada) Workstation 48 GB 48 GB 960 GB/s 1
AMD Ryzen AI Halo Developer Platform (Max+ 395) Mini PC 128 GB 64 GB 256 GB/s —
ArsenalPC MES2X Dual RTX 5090 AI Workstation Workstation 64 GB 64 GB — 2
BOSGAME M5 AI Mini Desktop (Ryzen AI Max+ 395) Mini PC 128 GB 64 GB — —
Chuwi UniBox AI395 (Ryzen AI Max+ 395) Mini PC 128 GB 64 GB 256 GB/s —
GMKtec EVO-X2 (Ryzen AI Max+ 395) Mini PC 128 GB 64 GB 256 GB/s —
MSI EdgeMesa N AI+ (RTX Spark N1X) Mini PC 128 GB 64 GB — —
MinisForum MS-S1 MAX (Ryzen AI Max+ 395) Mini PC 128 GB 64 GB 256 GB/s —
ACEMAGIC F9A (Ryzen AI Max+ PRO 495) Mini PC 192 GB 96 GB — —
BOXX APEXX 8R (1x RTX PRO 6000 Blackwell) Workstation 96 GB 96 GB 1792 GB/s 1
Chuwi UniBox AI495 Pro (Ryzen AI Max+ PRO 495) Mini PC 192 GB 96 GB 273 GB/s —
GEEKOM A9 Mega AI Workstation Workstation 128 GB 96 GB — 1
GMKtec EVO-X5 Pro (Ryzen AI Max+ PRO 495) Mini PC 192 GB 96 GB — —
NOVATECH RTX PRO 6000 AI Workstation Workstation 96 GB 96 GB — 1
System76 Thelio Major (1x RTX PRO 6000 Blackwell) Workstation 96 GB 96 GB 1792 GB/s 1
System76 Thelio Major (2x RTX 6000 Ada) Workstation 96 GB 96 GB 960 GB/s 2
Lenovo ThinkStation PGX Workstation 128 GB 115.2 GB 273 GB/s 1
NVIDIA DGX Spark Workstation 128 GB 115.2 GB 273 GB/s 1
BIZON G3000 G2 (4x RTX 5090) Workstation 128 GB 128 GB 1792 GB/s 4
Apple Mac Studio M2 Ultra (192GB) Workstation 192 GB 144 GB 800 GB/s 1
HP Z8 Fury G5 (4x RTX 6000 Ada) Workstation 192 GB 192 GB 960 GB/s 4

Frequently asked questions