Expired50% confidenceBenchmarkExact time
On Terminal Bench 2.1, the single-agent score is 88.8%, which reaches 91.9% with Ultra enabled, but the cost is approximately 3 times that of single-agent.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/15/2026
First Seen
Valid until: 10/13/2026(expired)
Sources
Related Claims
UnverifiedThe token consumption of concurrent multi-agent calls may in practice reach 3–8 times that of a single agent.59% similarUnverifiedBased on Anthropic's official pricing, cache-hit tokens are typically billed at approximately one-tenth the cost of normal input tokens.59% similarUnverifiedClaude Opus 4.8 achieved a completion rate of only 20.6% on OSWorld 2.0 with a 500-step budget, a partial score of 54.8%, and an average cost of approximately $72.4 per task.59% similarUnverifiedThe cost of the three-Agent architecture is approximately 20 times that of the single Agent ($200 vs. $9), but the output quality improved from 'unusable' to 'actually usable'59% similarUnverifiedBased on GPT-4o's input pricing of approximately $2.5 per million tokens, the annual cost of using AgentMemory is estimated to be only about 10 RMB.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/511328API
curl https://kongchang.com/api/v1/knowledge/claims/511328MCP
get_claim(id=511328)