Hardware for local LLMs: specs, tok/s, SG prices

Official specs and measured tok/s for AMD Strix Halo, NVIDIA DGX Spark, Apple Mac Max and Ultra (with Singapore prices), and GPU cards.

Snapshot 2026-10-16 · 394 configs on AI boxes and 423 on GPU cards · 123 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.

AI boxes

• AMD Strix Halo (Ryzen AI Max+ 395) · AMD Ryzen AI Max+ 395 (Strix Halo) · 178 single-stream configs · from S$4,321 in Singapore today (where to buy)
• AMD Gorgon Halo (Ryzen AI Max+ PRO 495) · AMD Ryzen AI Max+ PRO 495 (Gorgon Halo) · 0 single-stream configs
• NVIDIA DGX Spark (GB10) · NVIDIA GB10 Grace Blackwell Superchip · 128 single-stream configs · from S$6,624 in Singapore today (where to buy)
• Apple M-series Max (MacBook Pro / Mac Studio Max) · Apple M-series Max (MacBook Pro / Mac Studio Max) · 29 single-stream configs · from S$3,499 in Singapore today (where to buy)
• Apple Mac Studio (M-series Ultra) · Apple M-series Ultra — Mac Studio · 10 single-stream configs · from S$7,999 in Singapore today (where to buy)

GPU cards

GPU cards run the model from their own VRAM: very fast on what fits, but most cannot load what a 128 GB AI box can. Ranked on their own, never against the boxes.

• NVIDIA RTX 5090 (32GB) · 207 single-stream configs
• NVIDIA H100 PCIe (80GB) · 19 single-stream configs
• NVIDIA H100 SXM (80GB) · 15 single-stream configs
• NVIDIA A100 40GB · 19 single-stream configs
• NVIDIA A10 (24GB) · 8 single-stream configs

More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.