Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

End-to-end Attention Latency on 120K Context

62.9Attention Latency (120K Context, µs)

MAC-Attn.

-8.808475.221959.251,443.279Mar 31, 2026
Updated 16d ago

Evaluation Results

MethodLinks
2026.03
62.9
2026.03
64
2026.03
78.1
2026.03
103.8
2026.03
192.4
2026.03
210.8
2026.03
234.2
2026.03
234.2
2026.03
234.2
2026.03
234.2
2026.03
265.4
2026.03
324.6
2026.03
581.2
2026.03
594.7
2026.03
608.5
2026.03
640.5
2026.03
822.8
2026.03
844.7
2026.03
1,042.5
2026.03
1,855.6