[jev-gpu 2026-09-26T14:11:08Z] READY after 255s
[jev-gpu 2026-09-26T14:11:08Z] memory: Model loading took 14.8 GiB memory and 55.495697 seconds GPU KV cache size: 17,990 tokens, Maximum concurrency for 2,048 tokens per request: 8.78x
[jev-gpu 2026-09-26T14:16:23Z] load: 393.7 tok/s, range 389.5 to 396.3