Chips › Groq LPU (GroqCloud): specs, runtimes, measured speeds

Groq LPU (GroqCloud): specs, runtimes, measured speeds

Groq LPU (GroqCloud) specs with sources and dates, the runtimes that run on it with evidence, and the configs we measured.

Snapshot 2026-10-15 · 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.

Groq · cloud-api. 0 measured configs. Runtimes with an entry: 0.

Specs

• Memory: not stated · no source (2026-10-07)
• Memory type: not stated · no source (2026-10-07)
• Memory bandwidth: not stated · no source (2026-10-07)
• Peak AI compute (vendor figure): not stated · source (2026-10-07)
• Power (TDP): not stated · no source (2026-10-07)
• Process node: not stated · no source (2026-10-07)
• Launched: not stated · no source (2026-10-07)
• Status: cloud-only · source (2026-10-07)
• Price: $0.15-0.60 per million tokens (GPT-OSS 120B/20B) · source (2026-10-07)
• Rack compute, FP8 (vendor figure): 315 PFLOPS · source (2026-10-07)
• On-chip SRAM: 128 GB per rack · source (2026-10-07)

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