On February 15th, silicon-based mobility partnered with Beijing Ascend Artificial Intelligence Computing Center to support the private cluster deployment of hundreds of mainstream models based on Ascend’s computing power, including DeepSeeker R1&V3 series models, Qwen, FLUX. 1, etc., helping enterprise customers achieve intelligent upgrades. Specifically, based on the 16 card Ascend Atlas 800I A2, silicon-based mobility provides a single exclusive instance of DeepSeek-R1&V3 with full health, supporting over 100 concurrent instances; Based on the Ascend Atlas 800I A2, silicon-based flow supports adaptation to the 70B, 32B, and 14B distillation versions of DeepSeeker R1; Based on the Ascend Atlas 300I Duo, silicon-based flow supports adaptation to the 14B, 8B, and 7B distillation versions of DeepSeeker R1; Based on the Ascend Atlas 300V, silicon-based flow supports 1.5B adaptation of DeepSeek R1.
Scan code to share