A hosted inference platform for open-source and fine-tuned LLMs, aimed at teams that want fast, low-latency API access to those models without running their own GPU cluster.