Unverified50% confidenceFactExact time
The officially advertised 7x+ speedup primarily refers to TPS (tokens per second), i.e., throughput improvement in the generation phase, and does not necessarily mean an equivalent reduction in TTFT.
1
Sources
50%
Confidence
Long-term
Relevance
9/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHigh aggregate throughput often comes at the cost of larger batch sizes, and large batches drive up the time-to-first-token (TTFT) latency for individual requests.71% similarUnverifiedA faster first-token response (lower TTFT) creates the illusion of overall smoother performance, even if the total generation speed has not substantially improved.70% similarUnverifiedtokens/s指标可细分为首token延迟（TTFT）和生成速率（Generation Throughput）两个子维度65% similarUnverifiedIn some settings, the early reward climbing speed during RL when using TailSFT was 2.5 times faster than the standard SFT baseline.62% similarUnverifiedLLMOps核心指标包括每千Token成本、首Token延迟（TTFT）、提示词命中率和幻觉发生频率61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/960866API
curl https://kongchang.com/api/v1/knowledge/claims/960866MCP
get_claim(id=960866)