Frank
a9d45e7012
frank-archon: the slab is a threshold, and decode is where slots cost
...
Measured, 512 tokens, greedy, cold, at the deployed 3 x 200,000 fp8 shape:
prefill 22,208 -> 3,397, 24,000 -> 3,290, 25,600 -> 3,290 t/s at 43.5k —
the third slab step bought nothing, so the slab is a threshold whose cliff is
the 2026-10-11 arm at 26,071 (4,379), not a slope. Decode ladder: 99.0 t/s
single-stream at 43.5k, 93.3 at 197k (nearly window-free), and 3 concurrent
streams 113.9/43.3/10.9 = 18.4 t/s aggregate — the slot count is paid in
decode concurrency, the axis the shape hunt never measured because probe leg
B generates 24 tokens. probe-frank-archon-kv.py gains leg D (N-stream decode,
512 tokens, per-stream + aggregate) so no resize is priced on prefill alone.
2026-10-13 20:51:32 -07:00
..
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-06 22:41:18 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-08-28 02:20:06 -07:00
2026-09-16 21:43:09 -07:00
2026-09-20 17:51:27 -07:00
2026-10-10 17:03:14 -07:00
2026-09-16 06:59:44 -07:00
2026-10-09 16:47:19 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-09-16 15:38:20 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-09-21 18:56:35 -07:00
2026-09-20 15:25:09 -07:00
2026-10-10 17:03:14 -07:00
2026-10-09 16:47:19 -07:00
2026-09-16 07:43:33 -07:00
2026-10-07 21:31:10 -07:00
2026-09-20 15:25:09 -07:00
2026-10-10 17:03:14 -07:00
2026-10-09 16:47:19 -07:00
2026-08-25 20:13:33 -07:00
2026-08-25 20:13:33 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-08-25 20:13:33 -07:00
2026-10-10 17:03:14 -07:00
2026-10-06 21:42:16 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-09-26 19:51:57 -07:00
2026-10-09 16:47:19 -07:00
2026-10-10 17:03:14 -07:00
2026-10-13 20:33:02 -07:00
2026-10-13 19:20:33 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-08-28 02:20:06 -07:00
2026-09-26 15:23:24 -07:00
2026-09-26 22:50:17 -07:00
2026-10-05 23:26:51 -07:00
2026-09-21 13:20:15 -07:00
2026-08-23 12:59:56 -07:00
2026-08-23 12:59:56 -07:00
2026-10-10 17:03:14 -07:00
2026-10-07 02:45:28 -07:00
2026-10-10 17:03:14 -07:00
2026-08-25 15:38:57 -07:00
2026-10-13 20:51:32 -07:00
2026-10-05 23:26:51 -07:00
2026-09-26 19:25:46 -07:00
2026-10-10 17:03:14 -07:00
2026-09-26 23:25:55 -07:00
2026-10-13 20:18:41 -07:00
2026-10-10 17:03:14 -07:00
2026-10-10 17:03:14 -07:00
2026-09-17 18:38:54 -07:00
2026-09-17 13:09:59 -07:00
2026-10-06 00:21:40 -07:00
2026-10-10 17:03:14 -07:00
2026-08-25 17:04:15 -07:00
2026-10-13 19:20:33 -07:00
2026-10-06 20:41:33 -07:00