GLM 5.3 Flash on 4x 8 GB/Confirmation 4
From CMP Wiki
More actions
| Confirms | GLM 5.3 Flash on 4x 8 GB |
|---|---|
| Result | Reproduced |
| By | @ikyo |
| Date | 2026-10-06 |
| Cards | 8 GB |
| Numbers | Concurrency 1: 161.8 tok/s per user. 2: 231.0 aggregate. 4: 307.5 aggregate. 8: 296.1 aggregate, 4.55 s median TTFT. One user by workload: structured 182.1, code 161.8, prose 141.2 (768-token answers, reasoning counted as output). |
First run on a new 4-card box through Docker, on an early release Morrowmake said had bugs (a fix followed the same night). Source: @ikyo, 2026-10-06