Hardware to run Kimi K2.7 Code 1T (MoE)

Jun 2026 release — code-specialized variant built on K2.6 with ~30% fewer thinking tokens for the same task. Same 1 T total / 32 B active MoE (384 experts top-8 + 1 shared), 256 K context, Modified MIT. Internal Moonshot benchmarks only (Kimi Code Bench v2 62.0, MCP Atlas 76.0, MCP Mark Verified 81.1) — standard SWE-Bench / TB2 cells stay null until third-party leaderboards land.

Quantization Q2 Q2_K Q4 Q4_K_M Q5 Q5_K_M Q8 Q8_0

Availability ⚡ Only plug & play

Cheapest

8× Strix Halo cluster (1024 GB unified)

AMD · rack of 8 mini-PCs, 10/25 GbE fabric

$23,200

tokens / secQ4

235B-MoE 32 t/s

671B-MoE 8.0 t/s

1T-MoE 5.0 t/s

Memory1024 GB · 768 usable

Bandwidth256 GB/s

Idle / Active64 W / 960 W

Sticker$23,200

Why: Lowest sticker that still fits Kimi K2.7 Code 1T (MoE) ($23k USD).

Amazon (Strix Halo Mini PC) ↗GMKtec EVO-X2 ↗Bosgame M5 ↗Framework ↗

📺 Reviews on YouTube

▶ Kimi-K2 (1T) / GLM 4.7 (355B) on a 4-node Strix Halo cluster — closest documented baseline ▶ AMD Strix Halo / Ryzen AI Max+ 395 — honest review (single-node baseline)

Fastest

DGX B200 — 8× B200 server (1.44 TB HBM3e)

NVIDIA · 10U DGX server

$475,000

tokens / secQ4

235B-MoE 165 t/s

671B-MoE 105 t/s

1T-MoE 90 t/s

Memory1440 GB · 1404 usable

Bandwidth8000 GB/s

Idle / Active900 W / 10200 W

Sticker$475,000

Why: Highest measured tg/s — 90 t/s on Kimi K2.7 Code 1T (MoE)-class models at Q4.

NVIDIA DGX B200 ↗SuperMicro HGX B200 ↗Lambda Labs ↗

📺 Reviews on YouTube

▶ Inside a 1.44 TB HBM3e NVIDIA HGX B200 AI Server from ASRock Rack

All-rounder

2× Mac Studio M3 Ultra 512 GB cluster (TB5 / MLX)

Apple · two desktops, Thunderbolt 5 RDMA

$28,400

tokens / secQ4

235B-MoE 29 t/s

671B-MoE 9.0 t/s

1T-MoE 7.0 t/s

Memory1024 GB · 960 usable

Bandwidth819 GB/s

Idle / Active24 W / 440 W

Sticker$28,400

Why: Top quartile across speed, value, memory headroom, and efficiency — the "buy this if unsure" pick.

eBay ↗Apple Refurbished ↗B&H Photo (used) ↗

📺 Reviews on YouTube

▶ Apple didn't have to go this hard… (4× M3 Ultra TB5 cluster) ▶ Mac Studio CLUSTER vs M3 Ultra

Best value

8× RTX Pro 6000 Blackwell server (768 GB)

NVIDIA · 4U server (e.g. SuperMicro AS-4125GS-TNRT)

$78,000

tokens / secQ4

235B-MoE 220 t/s

671B-MoE 75 t/s

1T-MoE 40 t/s

Memory768 GB · 744 usable

Bandwidth1792 GB/s

Idle / Active220 W / 4800 W

Sticker$78,000

Why: Best $/tg-per-second — ~$1,950 per t/s.

Amazon ↗PNY Pro ↗CoreWeave ↗SuperMicro ↗

Best CUDA

DGX H200 — 8× H200 server (1.13 TB HBM3e)

NVIDIA · 8U DGX / HGX server rack

$380,000

tokens / secQ4

235B-MoE 155 t/s

671B-MoE 100 t/s

1T-MoE 85 t/s

Memory1128 GB · 1100 usable

Bandwidth4800 GB/s

Idle / Active700 W / 6500 W

Sticker$380,000

Why: Strongest CUDA-only software stack among fitting builds.

NVIDIA DGX H200 ↗SuperMicro HGX ↗Lambda Labs ↗

📺 Reviews on YouTube

▶ Inside the SUPER NVIDIA H200 Server From Supermicro (8U HGX H200)

Most VRAM

12× RTX Pro 6000 Blackwell rack (1152 GB)

NVIDIA · 8U server rack (multi-node, 1-2 chassis)

$118,000

tokens / secQ4

235B-MoE 260 t/s

671B-MoE 95 t/s

1T-MoE 55 t/s

Memory1152 GB · 1116 usable

Bandwidth1792 GB/s

Idle / Active340 W / 7400 W

Sticker$118,000

Why: 1116 GB usable — most headroom for batching and longer contexts.

Amazon ↗CoreWeave ↗SuperMicro ↗SHI enterprise ↗

Efficient

8× DGX Spark cluster (1024 GB unified, CUDA)

NVIDIA · rack of 8 desktops, 200 GbE fabric

$43,500

tokens / secQ4

235B-MoE 42 t/s

671B-MoE 16 t/s

1T-MoE 13 t/s

Memory1024 GB · 976 usable

Bandwidth273 GB/s

Idle / Active220 W / 1840 W

Sticker$43,500

Why: 1840 W active — lowest power draw of the fitting builds.

Amazon (8× DGX Spark) ↗Micro Center ↗NVIDIA Marketplace ↗ServeTheHome guide ↗

📺 Reviews on YouTube

▶ NVIDIA didn't want me to do this — 8× DGX Spark 1 TB VRAM cluster (AZisk)

No plug-and-play build fits at Q4_K_M

Only used / DIY / homelab-cluster rigs fit Kimi K2.7 Code 1T (MoE) at this quant. Turn off "Only plug & play" to see them.

Every other build that runs Kimi K2.7 Code 1T (MoE)

1 additional build fit Kimi K2.7 Code 1T (MoE) at Q4_K_M (600 GB usable minimum), sorted by sticker price.

Build	Price	Memory	Bandwidth	tg/s (Q4)	Active W	5-yr power
8× H100 80 GB serverNVIDIA · server rack	$280k	640 / 620 GB	3350 GB/s	60 t/s	5600 W	$20k

Last updated 2026-06-13