Compare
Compare local AI hardware with workload-aware output.
AMD Instinct MI350X 288GB wins for coding in balanced mode
Based on model fit, speed, and quality across top recommendations.
AMD Instinct MI350X 288GB
WinnerSDeepSeek V4 Flash
llama.cppq4-k-mRuns well
204.3 GB / 288.0 GB
110.0 tok/s1.0M ctx
SDevstral 2 123B Instruct
llama.cppq8-0Runs well
166.7 GB / 288.0 GB
53.1 tok/s256K ctx
SMiniMax M2.7
llama.cppq6-kRuns well
222.1 GB / 288.0 GB
105.6 tok/s205K ctx
Quick comparison
| Metric | AMD Instinct MI350X 288GB | RTX 4090 24GB |
|---|---|---|
| Models that fit | 3 | 3 |
| Avg decode tok/s | 89.6 | 37.9 |
| Best grade score | 99 | 93 |
