AI Inference. Blazing Fast. Better Economics (Sponsor)
AI agents make thousands of LLM calls. SambaNova's SN50 RDUs accelerate decoding, helping agents finish tasks faster with superior inference economics.
- GLM-5.3 live on SambaCloud at 200+ tokens/sec: Z.ai's model delivers state-of-the-art results on Terminal-Bench 3.0 and Agents' Last Exam, using fewer tokens per task than closed frontier models.
- Supermicro to build and sell SN50 systems: NVIDIA GPUs handle prefill; SN50 RDUs handle decode, with 10,000 chips initially.
- High performance, lower power demands: Run large models without extreme rack densities or new data centers.
