Compare
Compare local AI hardware with workload-aware output.
H100 NVL 188GB wins for coding in balanced mode
Based on model fit, speed, and quality across top recommendations.
AMD Instinct MI300X 192GB
SDevstral 2 123B Instruct
llama.cppq6-kRuns well
126.3 GB / 192.0 GB
46.8 tok/s212K ctx
SQwen 3.5 122B A10B
llama.cppq6-kRuns well
122.6 GB / 192.0 GB
129.8 tok/s131K ctx
SMistral Small 4 119B
llama.cppq6-kRuns well
123.1 GB / 192.0 GB
140.8 tok/s221K ctx
Quick comparison
| Metric | AMD Instinct MI300X 192GB | H100 NVL 188GB |
|---|---|---|
| Models that fit | 3 | 3 |
| Avg decode tok/s | 105.8 | 134.5 |
| Best grade score | 98 | 99 |