Unverified50% confidenceBenchmarkExact time
On the SemiAnalysis AgentX benchmark, the model achieves single-card throughput of 830,000 tokens/s (including cache, input and output), based on an interaction speed of 100 tokens/s per user.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/17/2026
First Seen
Valid until: 10/31/2026
Sources
Related Entities
Related Claims
UnverifiedA single agentic workflow task can produce tens of thousands of output tokens or more, whereas a single interaction in a traditional chat scenario often produces only a few hundred output tokens.64% similarUnverifiedThe token consumption of Agent workflows can reach 5 to 1000 times that of a simple query.64% similarUnverifiedProgressive tool discovery can cut the tokens resident in context from thousands down to a few hundred.62% similarUnverifiedAt peak usage, that user consumed over 700 million tokens in a single day.62% similarUnverified从个人代理转向团队代理的最大挑战之一是记忆管理，在100个用户场景下记忆消耗速度可能快上100倍62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1052495API
curl https://kongchang.com/api/v1/knowledge/claims/1052495MCP
get_claim(id=1052495)