← All chips

AI accelerator · China

Kunlunxin R480 / R300 (Kunlun Gen-2)

Second-generation Kunlun XPU-R accelerator family (2021), sold as the R200 full-height PCIe card, the R300 OAM compute module, and the R480-X8 server baseboard carrying 8 OAM modules for large-model training and inference clusters.

ChinaShippingKunlunxin (Baidu)Kunlun familyLaunched 2021-087 nm

Pricing not publicly disclosed · researched 2026-08-27

Specifications

Sourced specs

SpecValueSource
ArchitectureKunlun XPU-R (second generation)via kunlunxin.com2026-08
INT8 compute (per OAM module)256 TOPS x8 modulesvia kunlunxin.com2026-08
FP16 compute (per OAM module)128 TFLOPS x8 modules (INT16 128 TOPS, FP32 32 TFLOPS)via kunlunxin.com2026-08
Memory (per OAM module)32 GB GDDR6, 512 GB/s bandwidthvia kunlunxin.com2026-08
Interconnect200 GB/s (chip-to-chip)via kunlunxin.com2026-08
Process node7 nmvia kunlunxin.com2026-08
Form factorOCP-OAI standard: UBB baseboard with 8 OAM modules (R480); R200 PCIe card and R300 OAM module variantsvia finance.sina.cn2021-12-14

Pricing

Price history, newest first

No public price observation yet.

vendor-official or multiple independent reports · single major report or several secondary · estimate or lone secondary source

Prices in USD, converted at each observation year's average exchange rate; the original figure and rate print under every converted price. How prices are sourced and graded ↗

Context

Notes

  • The variant name 'R480-8000' does not appear on Kunlunxin's official site; the current 8-OAM product is the R480-X8.
  • No public price or cloud rental rate located for R200/R300/R480; sales are B2B, and Baidu Cloud GPU instance pricing requires login.
  • Kunlun II (XPU-R) generation was announced August 2021 with mass production following; single R480-X8 system delivers up to 1 PFLOPS FP16 per Tencent Tech.

Where it sits

In the stack

All specs and prices link to their original sources, dated as observed. Figures are market observations, not offers; treat each as of its date.