Microsoft’s Maia 200 inference AIP Triples computational throughput, reduces inference latency and costs per token.
