ajensenwaud commited on
Commit
dc7ebc4
·
verified ·
1 Parent(s): eb65d8b

Model card: speed from the release build

Browse files
Files changed (1) hide show
  1. README.md +5 -5
README.md CHANGED
@@ -43,12 +43,12 @@ Global-MMLU-Lite lands inside its ±2.7-point interval of the card's figure.
43
 
44
  | | 1M-token window | 8 x 128K windows |
45
  |---|---|---|
46
- | Decode at 128K context, plain / with the drafter | 24.9 / 36.8 tok/s | 22.9 / 38.9 tok/s (120K) |
47
- | Decode at 1M context, plain / with the drafter | 9.4 / 17.0 tok/s | |
48
- | Reading a cold 64K prompt | | 717 tok/s (91 s) |
49
- | Eight streams at once | | 92 tok/s in total |
50
 
51
- A cold 1M-token prompt takes most of an hour to read; follow-up turns resume from the prompt cache in seconds.
52
 
53
  ## Use
54
 
 
43
 
44
  | | 1M-token window | 8 x 128K windows |
45
  |---|---|---|
46
+ | Decode at 128K context, plain / with the drafter | 25.9 / 37.9 tok/s | 23.3 / 36.9 tok/s (120K) |
47
+ | Decode at 1M context, plain / with the drafter | 13.3 / 21.1 tok/s | |
48
+ | Reading a cold 64K prompt | | 774 tok/s (84 s) |
49
+ | Eight streams at once | | 89-90 tok/s in total |
50
 
51
+ A cold 1M-token prompt takes about 46 minutes to read; follow-up turns resume from the prompt cache in seconds.
52
 
53
  ## Use
54