Unverified50% confidenceConceptExact time
tps (tokens per second) is a core metric for measuring model inference speed: below 30 tps you will clearly feel the wait, 60-100 tps can satisfy most dialogue and coding needs, and above 100 tps is quite smooth.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedTPS measurement is typically divided into two phases: Time to First Token (TTFT) and the generation rate (throughput) of subsequent tokens.78% similarUnverifiedWhen a model's generation speed exceeds 15–20 TPS, users no longer perceive noticeable waiting during interaction, because humans read English at approximately 4–5 tokens per second74% similarUnverifiedAn agent task requiring 20 LLM calls with 200 tokens generated each time takes approximately 133 seconds to complete at 30 TPS, but only 13 seconds at 300 TPS.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1017392API
curl https://kongchang.com/api/v1/knowledge/claims/1017392MCP
get_claim(id=1017392)