Unverified50% confidenceBenchmarkExact time
JetBrains testing showed that Caveman claimed to save 65% of tokens but actually cut only about 8.5% of output tokens; RTK saved essentially nothing and in low-effort mode actually made tasks more expensive; Ponytail saved about 10% per task, half its claimed value.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/8/2026
First Seen
Valid until: 11/22/2026
Sources
We Tested 100+ Claude GitHub Repos — These 6 Are Worth Installing
youtubeNate Herk | AI Automation10/11/2026
Related Entities
Related Claims
UnverifiedRTK compressed tool characters by 9%, but prompt tokens actually increased by 9%, with average per-round input down 7% but total rounds up 18%.66% similarUnverifiedWhen the models were configured as agents with a streamlined skill and a checker tool, the valid tok generation rate rose to 97.6% to 100%, and the token savings for the strongest model were 21% and 14%.66% similarUnverifiedThe 35%–55% token savings are strongly correlated with the task distribution; the more tasks skew toward simple and repetitive, the more pronounced the gains.61% similarUnverifiedFreeToken supports runtime adjustment of the GPU expert cache size; tests show that reducing the cache from 58% to 40% results in only about 10% performance loss, with performance dropping sharply below 20%.61% similarUnverifiedIn the SWE BenchPro test, Opus with medium solved 92.8% of tasks and Fable with high solved 92.3% of tasks, but Fable's cost per solved task was about 5 times that of the former.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1147962API
curl https://kongchang.com/api/v1/knowledge/claims/1147962MCP
get_claim(id=1147962)