Google AI ModeSep 18, 2026
MNN / TensorRT / OpenVINO (Intel): While often used for off-the-shelf chips, startups write custom backends or plugins for these to benchmark latency/energy consumption profiles during early-stage architectural exploration.
