Unverified50% confidenceBenchmarkExact time
On a Blackwell workstation GPU, the acceptance rate is 81% with 2 draft tokens, 56% with 5, and 40% with 8.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/10/2026
First Seen
Valid until: 12/9/2026
Sources
Related Entities
Related Claims
UnverifiedThroughput peaks at approximately 115 tokens per second around 5 draft tokens and then falls back, not varying linearly with the acceptance rate.62% similarUnverifiedIn empirical tests, the draft token acceptance rate of DFlash2 is generally between 40% and 50%, reaching approximately 58%–60% at its higher end.61% similarUnverifiedIt is recommended to use the Compact command when context usage reaches 60%–80%.61% similarUnverifiedOn the Terminal Bench 2.1 benchmark, Ultra mode approaches 92%, surpassing competitors' performance of 88%.60% similarUnverifiedAfter intelligent filtering, the draft acceptance rate in mathematical reasoning scenarios increased from 76.9% to 92.5%.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/887332API
curl https://kongchang.com/api/v1/knowledge/claims/887332MCP
get_claim(id=887332)