Expired50% confidenceBenchmarkExact time
In the long-task test, the plugin version's average context was about 210,000 Tokens, about 40% less than the original; the ending context was 320,000 Tokens, about 25% less than the original.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/4/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 11/18/2026(expired)
Sources
Open-Source Analysis of Live-Compact, the Real-Time Compression Plugin for DeepSeek Harness
bilibiliMerkyor11/3/2026
Related Entities
Related Claims
UnverifiedFor most application scenarios, the operating cost of a well fine-tuned 27B–32B model is only one-tenth that of a 400B+ level model.70% similarUnverifiedgpt-oss-120b spent about 25 times the cost per correct answer on the distractor counting task, while its accuracy was 7.5 percentage points lower (not significant).68% similarUnverifiedOn 36 small verification tasks, Opus 5.5 used about 40% fewer tokens than Opus at the halfway point.67% similarUnverified对于10万行代码的项目,使用向量检索每次任务可能只需加载2%~5%的相关代码,Token消耗可降低20倍以上67% similarUnverifiedOn July 10, 2025, METR published a randomized controlled trial: 16 senior open-source developers handled 246 tasks on real repositories, and after using AI tools they were actually 19% slower, despite prior expectations of a 24% speedup and post-hoc self-reported estimates of a 20% speedup.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133448API
curl https://kongchang.com/api/v1/knowledge/claims/1133448MCP
get_claim(id=1133448)