Speed demo: Glm5.3 flash on 2x GB10s
https://relay.utxo.one/8125e248f9192acc5eb8cfcb812e6708c4eefdcedde6bdb1f885626180ecd779.mp4
Yes I'm similar. About 50 tps for code and 25 tps for reasoning which is where u feel the lack of memory bandwidth. I got spoiled with Deepseek v4 flash 0731 speed