List your company

End-to-end Time-to-First-Token (TTFT) on Long-context sequences

257TTFT (ms)FlashPrefill
-1,882.812,560.8527,004.541,448.15Mar 6, 2026

Evaluation results

MethodLinks
FlashPrefillModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=4K2026.032571.04
FullModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=4K2026.03267—
FlashPrefillModel=Qwen2.5-7B-Instruct, Sequence Length=4K2026.034061.04
FullModel=Qwen2.5-7B-Instruct, Sequence Length=4K2026.03423—
FlashPrefillModel=Llama-3.1-8B-Instruct, Sequence Length=4K2026.034471.05
FullModel=Llama-3.1-8B-Instruct, Sequence Length=4K2026.03471—
FlashPrefillModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=8K2026.035191.18
FullModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=8K2026.03613—
FlashPrefillModel=Qwen2.5-7B-Instruct, Sequence Length=8K2026.038511.07
FullModel=Qwen2.5-7B-Instruct, Sequence Length=8K2026.03912—
FlashPrefillModel=Llama-3.1-8B-Instruct, Sequence Length=8K2026.039131.07
FullModel=Llama-3.1-8B-Instruct, Sequence Length=8K2026.03976—
FlashPrefillModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=16K2026.031,0901.47
FullModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=16K2026.031,603—
FlashPrefillModel=Qwen2.5-7B-Instruct, Sequence Length=16K2026.031,7071.2
FlashPrefillModel=Llama-3.1-8B-Instruct, Sequence Length=16K2026.031,8601.18
FullModel=Qwen2.5-7B-Instruct, Sequence Length=16K2026.032,048—
FullModel=Llama-3.1-8B-Instruct, Sequence Length=16K2026.032,189—
FlashPrefillModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=32K2026.032,3311.98
FlashPrefillModel=Qwen2.5-7B-Instruct, Sequence Length=32K2026.033,5341.34
FlashPrefillModel=Llama-3.1-8B-Instruct, Sequence Length=32K2026.033,6181.49
FullModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=32K2026.034,635—
FullModel=Qwen2.5-7B-Instruct, Sequence Length=32K2026.034,735—
FlashPrefillModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=64K2026.034,9093.1
FullModel=Llama-3.1-8B-Instruct, Sequence Length=32K2026.035,398—
FlashPrefillModel=Qwen2.5-7B-Instruct, Sequence Length=64K2026.037,3231.72
FlashPrefillModel=Llama-3.1-8B-Instruct, Sequence Length=64K2026.037,4811.97
FlashPrefillModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=128K2026.0310,7025.02
FullModel=Qwen2.5-7B-Instruct, Sequence Length=64K2026.0312,563—
FullModel=Llama-3.1-8B-Instruct, Sequence Length=64K2026.0314,774—
FlashPrefillModel=Llama-3.1-8B-Instruct, Sequence Length=128K2026.0315,0613.02
FullModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=64K2026.0315,203—
FlashPrefillModel=Qwen2.5-7B-Instruct, Sequence Length=128K2026.0315,2132.45
FullModel=Qwen2.5-7B-Instruct, Sequence Length=128K2026.0337,239—
FullModel=Llama-3.1-8B-Instruct, Sequence Length=128K2026.0345,464—
FullModel=Qwen3-30B-A3B-Instruct-2507, Sequence Length=128K2026.0353,752—