Unverified50% confidenceFactExact time
The industry generally considers a generation speed of around 10 tokens per second to be close to human reading speed, and reaching this level is enough to support everyday conversation and document processing.
1
Sources
50%
Confidence
Long-term
Relevance
11/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWith all three optimizations stacked, generation speed stabilizes at 9 to 10 tokens per second, but time-to-first-token for a 10,000-character long document still requires approximately 329 seconds.73% similarUnverifiedIn libraries with more than 10,000 items, Deepmark keeps search response time under 100 milliseconds.72% similarUnverifiedMicrosoft Research Asia's MInference framework matches the most suitable sparse computation method to different attention heads, achieving up to 10x speedup in the prefill stage of million-word long contexts without reducing retrieval accuracy.69% similarUnverifiedAfter modern tokenizers are highly optimized, the time for a single tokenization operation is typically on the order of 10–50 ms68% similarUnverifiedThe time spent reading code during software development far exceeds the time spent writing code, with the ratio typically above 10:1.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1150128API
curl https://kongchang.com/api/v1/knowledge/claims/1150128MCP
get_claim(id=1150128)