Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
Every figure here is computed, not measured. Fit is weights at each quantisation against device memory, with a 25% allowance for the KV cache, activations and the OS. Decode rate is the memory-bandwidth roofline at 70% efficiency. Nobody has run this model on these devices. This is a mixture-of-experts model and no card carries active-parameter counts, so these predictions use total parameters and understate the real speed.
Device
Memory
Bandwidth
Best quality
Weights
tok/s
Fastest
NVIDIA GB200 Grace Blackwell Superchip
372.0 GB
16000.0 GB/s
fp8
180.0 GB
~62.2
~124.4 q4
AMD Instinct MI355X
288.0 GB
8000.0 GB/s
fp8
180.0 GB
~31.1
~62.2 q4
NVIDIA B200
180.0 GB
7700.0 GB/s
q6
135.0 GB
~39.9
~59.9 q4
Google TPU7x (Ironwood)
192.0 GB
7380.0 GB/s
q6
135.0 GB
~38.3
~57.4 q4
AMD Instinct MI325X
256.0 GB
6000.0 GB/s
fp8
180.0 GB
~23.3
~46.7 q4
AMD Instinct MI300X
192.0 GB
5300.0 GB/s
q6
135.0 GB
~27.5
~41.2 q4
NVIDIA H200 SXM
141.0 GB
4800.0 GB/s
q4
90.0 GB
~37.3
~37.3 q4
AMD Instinct MI250X
128.0 GB
3200.0 GB/s
q4
90.0 GB
~24.9
~24.9 q4
Apple M3 Ultra
512.0 GB
819.0 GB/s
bf16
360.0 GB
~1.6
~6.4 q4
Apple M2 Ultra
192.0 GB
800.0 GB/s
q6
135.0 GB
~4.1
~6.2 q4
Apple M4 Max
128.0 GB
546.0 GB/s
q4
90.0 GB
~4.2
~4.2 q4
Apple M3 Max
128.0 GB
400.0 GB/s
q4
90.0 GB
~3.1
~3.1 q4
Where it runs
Taken from the card's availability section.
Platform
Id on that platform
Hugging Face
Qwen/Qwen3.8-Flash-Next
Verified benchmark evidence
Each row was checked against its source by a reviewer, and carries the model identifier as actually evaluated — which is not always the same as this card's.