Inference Performance Evaluation on GPU Context Length Sweep (2K-8K)
116TTFT
SpectrumKV
Updated 1mo ago
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
2026.06 | 116 | 58 | 50.1 | 3 | |
2026.06 | 135 | 50 | 38.9 | 2 | |
2026.06 | 154 | 61 | 43.5 | 11 | |
2026.06 | 233 | 53 | 49.1 | 4 | |
2026.06 | 269 | - | 38.2 | - | |
2026.06 | 270 | 62 | 38.2 | 33 | |
2026.06 | 274 | - | 51.9 | - | |
2026.06 | 296 | 53 | 43.2 | 2 | |
2026.06 | 391 | - | 49.2 | - | |
2026.06 | 497 | - | 47.1 | - | |
2026.06 | 502 | 59 | 47 | 7 | |
2026.06 | 630 | - | 42.4 | - | |
2026.06 | 709 | - | 28.8 | - | |
2026.06 | 713 | 57 | 28.8 | 5 | |
2026.06 | 1,211 | - | 44.1 | - | |
2026.06 | 1,664 | - | 27.3 | - |