[Cerebras Launches CS-4 AI Accelerator, Claims Inference Speed Up to 30x Faster Than GPU Solutions] According to Jiemian News, Cerebras has unveiled its next-generation AI accelerator, the CS-4, claiming that its inference speed can reach up to 30 times that of GPU-based solutions. The CS-4 is powered by three new WSE-3 Turbo wafer-scale engines, delivering AI compute performance of 750 PFLOPS and memory bandwidth of 129.6 PB/s. Cerebras stated that in the GPT-OSS-120B test, the CS-4 generates over 4,400 tokens per second per user, achieving speeds up to 30 times faster than GPU solutions. Its throughput per watt is up to 10 times higher than the CS-3 and supports models with over 50 trillion parameters. The CS-4 is expected to begin shipping in the third quarter of this year.