Hardware to run Mistral Medium 3.5 128B

Apr 30 2026. Western 128B dense with vision + 256 K context. 77.6% SWE-Bench Verified; first credible mid-tier open-weight from Mistral in months. Modified MIT.

Quantization Q2 Q2_K Q4 Q4_K_M Q5 Q5_K_M Q8 Q8_0

Availability ⚡ Only plug & play

Cheapest

Quad AMD MI50 32 GB (128 GB) homelab build

AMD · rack/large tower

$2,500

tokens / secQ4

30B 50 t/s

70B 35 t/s

120B-MoE 45 t/s

Memory128 GB · 122 usable

Bandwidth1024 GB/s

Idle / Active75 W / 1200 W

Sticker$2,500

Why: Lowest sticker that still fits Mistral Medium 3.5 128B ($2.5k USD).

AmazoneBay (build 4× used)Alibaba

📺 Reviews on YouTube

▶ AMD MI50 32 GB for local AI — Qwen 3.6 & Gemma 4 on llama.cpp / vLLM ▶ Scalable local AI build — Qwen3.6-27B, Gemma4-31B, Qwen3.5-122B

Fastest

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

NVIDIA · 10U DGX server

$475,000

tokens / secQ4

30B 270 t/s

70B 180 t/s

120B-MoE 330 t/s

Memory1440 GB · 1404 usable

Bandwidth8000 GB/s

Idle / Active900 W / 10200 W

Sticker$475,000

Why: Highest measured tg/s — 98 t/s on Mistral Medium 3.5 128B-class models at Q4.

NVIDIA DGX B200SuperMicro HGX B200Lambda Labs

📺 Reviews on YouTube

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

All-rounder

Mac Studio M3 Ultra 256 GB

Apple · small desktop

$7,999

tokens / secQ4

30B 38 t/s

70B 18 t/s

120B-MoE 55 t/s

Memory256 GB · 232 usable

Bandwidth819 GB/s

Idle / Active10 W / 180 W

Sticker$7,999

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

AppleB&H Photo

📺 Reviews on YouTube

▶ M3 Ultra Mac Studio Review

Best value

Quad RTX 3090 (used) build

NVIDIA · rack/large tower

$6,000

tokens / secQ4

30B 40 t/s

70B 28 t/s

120B-MoE 100 t/s

Memory96 GB · 92 usable

Bandwidth936 GB/s

Idle / Active90 W / 1400 W

Sticker$6,000

Why: Best $/tg-per-second — ~$392 per t/s.

AmazoneBay (build 4× used)

📺 Reviews on YouTube

▶ ULTIMATE Local AI Quad 3090 Build

Best CUDA

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

NVIDIA · 8U DGX / HGX server rack

$380,000

tokens / secQ4

30B 250 t/s

70B 170 t/s

120B-MoE 310 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 H200SuperMicro HGXLambda 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

30B 250 t/s

70B 170 t/s

120B-MoE —

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.

AmazonCoreWeaveSuperMicroSHI enterprise

Efficient

MacBook Pro M5 Max 128 GB

Apple · laptop

$5,399

tokens / secQ4

30B 30 t/s

70B 14 t/s

120B-MoE 65 t/s

Memory128 GB · 108 usable

Bandwidth614 GB/s

Idle / Active6 W / 95 W

Sticker$5,399

Why: 95 W active — lowest power draw of the fitting builds.

AmazonAppleB&H Photo

📺 Reviews on YouTube

▶ Apple's New M5 Max Changes the Local AI Story (AZisk) ▶ I Linked 3 M5 MacBooks to Run "Impossible" AI (AZisk) ▶ I was WRONG about M5 MacBook Pro vs M4 Max & M3 Ultra for Local AI (xCreate)

Cheapest

AMD Ryzen AI Max+ 395 (128 GB)

AMD · mini desktop / laptop

$2,799

tokens / secQ4

30B 16 t/s

70B 5.0 t/s

120B-MoE 47 t/s

Memory128 GB · 96 usable

Bandwidth256 GB/s

Idle / Active8 W / 120 W

Sticker$2,799

Why: Lowest sticker that still fits Mistral Medium 3.5 128B ($2.8k USD).

AmazonGMKtec EVO-X2Framework

📺 Reviews on YouTube

▶ AMD Strix Halo / Ryzen AI Max+ 395 — an honest review ▶ Tuning the 128 GB AMD AI mini PC for fast inference ▶ Running vLLM on Strix Halo + ROCm performance updates

Fastest

Single B200 180 GB workstation

NVIDIA · workstation / 4U server

$47,000

tokens / secQ4

30B 135 t/s

70B 75 t/s

120B-MoE 120 t/s

Memory180 GB · 176 usable

Bandwidth8000 GB/s

Idle / Active100 W / 1000 W

Sticker$47,000

Why: Highest measured tg/s — 41 t/s on Mistral Medium 3.5 128B-class models at Q4.

NVIDIA B200 partnersSHI enterpriseLambda Labs

📺 Reviews on YouTube

▶ AI Lab: NVIDIA B200 vs GB200 explained | GPU architecture for LLMs ▶ Inside a NEW AI Cluster — Tour with NVIDIA B200

All-rounder

Mac Studio M3 Ultra 256 GB

Apple · small desktop

$7,999

tokens / secQ4

30B 38 t/s

70B 18 t/s

120B-MoE 55 t/s

Memory256 GB · 232 usable

Bandwidth819 GB/s

Idle / Active10 W / 180 W

Sticker$7,999

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

AppleB&H Photo

📺 Reviews on YouTube

▶ M3 Ultra Mac Studio Review

Best value

Mac Studio M3 Ultra 96 GB

Apple · small desktop

$5,299

tokens / secQ4

30B 38 t/s

70B 18 t/s

120B-MoE 55 t/s

Memory96 GB · 80 usable

Bandwidth819 GB/s

Idle / Active10 W / 180 W

Sticker$5,299

Why: Best $/tg-per-second — ~$538 per t/s.

AmazonAppleB&H Photo

📺 Reviews on YouTube

▶ M3 Ultra vs M4 Max — Don't Buy the WRONG Mac Studio ▶ Skip M3 Ultra & RTX 5090 for LLMs | NEW 96GB KING ▶ Ditch 512 GB Monster — this M3 Ultra Just Redefined 'Enough'

Best CUDA

Single H200 141 GB workstation

NVIDIA · workstation / 2U server

$40,000

tokens / secQ4

30B 125 t/s

70B 70 t/s

120B-MoE 112 t/s

Memory141 GB · 138 usable

Bandwidth4800 GB/s

Idle / Active60 W / 700 W

Sticker$40,000

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

NVIDIA H200 partnersSHI enterpriseLambda Labs

📺 Reviews on YouTube

▶ Nvidia Wouldn't Send Me This $30,000 GPU — H200 Holy $H!T ▶ Inside the SUPER NVIDIA H200 Server From Supermicro

Most VRAM

Mac Studio M3 Ultra 512 GB

Apple · small desktop

$14,199

tokens / secQ4

30B 38 t/s

70B 18 t/s

120B-MoE 55 t/s

Memory512 GB · 480 usable

Bandwidth819 GB/s

Idle / Active12 W / 220 W

Sticker$14,199

Why: 480 GB usable — most headroom for batching and longer contexts.

Apple RefurbB&H Photo (used)eBay (used)

📺 Reviews on YouTube

▶ M3 Ultra Mac Studio Review ▶ M3 Ultra vs RTX 5090 — The Final Battle

Efficient

MacBook Pro M5 Max 128 GB

Apple · laptop

$5,399

tokens / secQ4

30B 30 t/s

70B 14 t/s

120B-MoE 65 t/s

Memory128 GB · 108 usable

Bandwidth614 GB/s

Idle / Active6 W / 95 W

Sticker$5,399

Why: 95 W active — lowest power draw of the fitting builds.

AmazonAppleB&H Photo

📺 Reviews on YouTube

▶ Apple's New M5 Max Changes the Local AI Story (AZisk) ▶ I Linked 3 M5 MacBooks to Run "Impossible" AI (AZisk) ▶ I was WRONG about M5 MacBook Pro vs M4 Max & M3 Ultra for Local AI (xCreate)

Every other build that runs Mistral Medium 3.5 128B

30 additional builds fit Mistral Medium 3.5 128B at Q4_K_M (80 GB usable minimum), sorted by sticker price.

Build	Price	Memory	Bandwidth	tg/s (Q4)	Active W	5-yr power
Quad Tesla P40 (96 GB) homelab buildNVIDIA · rack/large tower	$2.7k	96 / 92 GB	347 GB/s	2.2 t/s	1000 W	$3.5k
ASUS Ascent GX10 (128 GB)ASUS · small desktop	$4.0k	128 / 119 GB	273 GB/s	4.4 t/s	240 W	$903
Dell Pro Max with GB10 (128 GB)Dell · small desktop	$4.1k	128 / 119 GB	273 GB/s	4.4 t/s	240 W	$887
Lenovo ThinkStation PGX (128 GB)Lenovo · small desktop	$4.1k	128 / 119 GB	273 GB/s	4.4 t/s	160 W	$650
MSI EdgeXpert MS-C931 (128 GB)MSI · small desktop	$4.7k	128 / 119 GB	273 GB/s	4.4 t/s	240 W	$887
Mac Studio M4 Max 128 GBApple · small desktop	$4.7k	128 / 112 GB	546 GB/s	7.1 t/s	130 W	$453
NVIDIA DGX Spark (128 GB)NVIDIA · small desktop	$4.7k	128 / 119 GB	273 GB/s	4.4 t/s	240 W	$887
2× Strix Halo cluster (256 GB unified)AMD · mini-PC pair	$6k	256 / 192 GB	256 GB/s	3.3 t/s	240 W	$841
Quad Intel Arc Pro B70 buildIntel · rack/large tower	$6k	128 / 124 GB	608 GB/s	12 t/s	700 W	$2.5k
Dual AMD Radeon Pro W7900 buildAMD · workstation	$9k	96 / 92 GB	864 GB/s	12 t/s	600 W	$2.1k
2× DGX Spark cluster (256 GB unified, CUDA)NVIDIA · two desktops, 200 G interconnect	$10k	256 / 240 GB	273 GB/s	7.7 t/s	460 W	$1.7k
Octuple Intel Arc Pro B70 clusterIntel · rack/large tower	$11k	256 / 248 GB	608 GB/s	15 t/s	1450 W	$5k
4× Strix Halo cluster (512 GB unified)AMD · rack of 4 mini-PCs, 10 GbE fabric	$12k	512 / 384 GB	256 GB/s	4.4 t/s	480 W	$1.7k
Single RTX Pro 6000 Blackwell 96 GB buildNVIDIA · workstation	$12k	96 / 93 GB	1792 GB/s	19 t/s	600 W	$2.1k
Tinybox Red (6× 7900 XTX, 144 GB)tinycorp · 12U pedestal	$15k	144 / 138 GB	960 GB/s	15 t/s	1500 W	$5k
4× DGX Spark cluster (512 GB unified, CUDA)NVIDIA · rack of 4 desktops	$20k	512 / 488 GB	273 GB/s	9.8 t/s	920 W	$3.4k
8× Strix Halo cluster (1024 GB unified)AMD · rack of 8 mini-PCs, 10/25 GbE fabric	$23k	1024 / 768 GB	256 GB/s	4.9 t/s	960 W	$3.4k
Dual RTX Pro 6000 Blackwell buildNVIDIA · workstation	$24k	192 / 188 GB	1792 GB/s	33 t/s	1100 W	$3.8k
Single AMD Instinct MI325X 256 GB workstationAMD · workstation / 4U server (OAM)	$25k	256 / 250 GB	6000 GB/s	—	1000 W	$3.6k
Tinybox Green (6× RTX 4090, 144 GB)tinycorp · 12U pedestal	$25k	144 / 138 GB	1008 GB/s	21 t/s	2200 W	$8k
Single AMD Instinct MI355X 288 GB workstationAMD · 4U server (OAM, liquid-cooled)	$28k	288 / 282 GB	8000 GB/s	—	1400 W	$5k
2× Mac Studio M3 Ultra 512 GB cluster (TB5 / MLX)Apple · two desktops, Thunderbolt 5 RDMA	$28k	1024 / 960 GB	819 GB/s	13 t/s	440 W	$1.5k
Single AMD Instinct MI300X 192 GB workstationAMD · workstation	$30k	192 / 188 GB	5300 GB/s	—	750 W	$2.8k
Quad RTX Pro 6000 Blackwell build (384 GB)NVIDIA · workstation / 4U pedestal	$38k	384 / 372 GB	1792 GB/s	52 t/s	2200 W	$8k
Tinybox Pro (8× RTX 4090, 192 GB)tinycorp · 12U pedestal	$40k	192 / 184 GB	1008 GB/s	27 t/s	3200 W	$11k
8× DGX Spark cluster (1024 GB unified, CUDA)NVIDIA · rack of 8 desktops, 200 GbE fabric	$44k	1024 / 976 GB	273 GB/s	13 t/s	1840 W	$7k
Tinybox Green v2 (4× RTX 5090, 128 GB)tinycorp · 12U pedestal	$45k	128 / 124 GB	1792 GB/s	33 t/s	2300 W	$8k
8× RTX Pro 6000 Blackwell server (768 GB)NVIDIA · 4U server (e.g. SuperMicro AS-4125GS-TNRT)	$78k	768 / 744 GB	1792 GB/s	77 t/s	4800 W	$16k
8× H100 80 GB serverNVIDIA · server rack	$280k	640 / 620 GB	3350 GB/s	66 t/s	5600 W	$20k
NVIDIA RTX Spark (128 GB)NVIDIA · OEM laptops + small desktops	—	128 / 119 GB	300 GB/s	—	— W	$756

Last updated 2026-07-11