Tsinghua + Moondark’s PrfaaS is quite interesting: long-context is broken out for pre-filling, short requests are locally decoded, bandwidth-aware scheduling avoids blowing up the link, and the measured data is solid too.

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.