Unverified50% confidenceBenchmarkExact time
In a controlled test using the same code repository, the same model, and the same workflow, Codex input tokens routed through the Edgee compression gateway were reduced by 49.5%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/18/2026
First Seen
Valid until: 12/17/2026
Sources
Related Entities
Related Claims
UnverifiedWork such as LLMLingua from Microsoft Research has demonstrated that removing redundant tokens can compress input tokens by 40%–60% while maintaining task accuracy above 95%.64% similarUnverifiedWhen prefix caching and DFlash2 speculative decoding are both enabled simultaneously, total throughput drops by approximately 37%, which is even slower than not enabling speculative decoding at all.64% similarUnverifiedCodeGraph tests show a reduction in Token consumption of approximately 47%.63% similarUnverifiedCodeGraph tests show a reduction in the number of tool calls of approximately 58%.62% similarUnverified混合架构中，仅当规则匹配失败或置信度低于阈值时才调用大模型兜底，通常可减少60%-80%的API调用量62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/932086API
curl https://kongchang.com/api/v1/knowledge/claims/932086MCP
get_claim(id=932086)