MLA Decode Latency on AMD MI300X Instance 3
1,668.7Decode Latency (µs)1st human
Evaluation Results
| Method | Links | |
|---|---|---|
| 1st humanModel=N/A2026.01 | 1,668.7 | |
| TTT-DiscoverModel=gpt-oss-120b2026.01 | 1,671.3 | |
| 2nd humanModel=N/A2026.01 | 1,679.7 | |
| 3rd humanModel=N/A2026.01 | 1,718 | |
| 4th humanModel=N/A2026.01 | 1,767 | |
| 5th humanModel=N/A2026.01 | 2,041.9 | |
| Best-of-25600Model=gpt-oss-120b2026.01 | 2,275.2 |