The preview version of DeepSeek-V4 was launched today (24th) and was simultaneously open-sourced, including DeepSeek-V4-Pro and DeepSeek-V4-Flash, according to STAR Market Daily. DeepSeek-V4 completed validation of its fine-grained expert parallelism (EP) solution on both Nvidia's GPUs and Huawei Ascend NPUs, according to DeepSeek's official document introduction. Moreover, the official API page noted that, constrained by high-end computing power, the service throughput of the V4-Pro model remains limited. Related NewsOutstanding Loan Growth YoY for Apr in China is 5.6%, lower than the previous value of 5.8%. The forecast was 5.8%.It is expected that, after the Ascend 950 SuperPod is launched in bulk in 2H26, the V4-Pro model price will be significantly reduced.CAMBRICON (688256.SH) -51.580 (-4.092%) announced on its WeChat official account that it has completed Day 0 adaptation of the latest open-source models from DeepSeek, naming 285B DeepSeek-V4-Flash and 1.6T DeepSeek-V4-Pro, based on the vLLM inference framework. The adaptation code has been open-sourced to the GitHub community.(A Shares quote is delayed for at least 15 mins.)
AASTOCKS Financial News