Unverified50% confidenceBenchmarkExact time
The number of tokens consumed by MATE is approximately one-tenth that of the original trajectory.
1
Sources
50%
Confidence
Long-term
Relevance
9/30/2026
First Seen
Sources
Related Entities
Related Claims
Unverified同一段文字在不同模型中消耗的 token 数量可能相差 10%-30%68% similarUnverifiedSol's output token count and completion time are both less than half of competitors', with an estimated cost approximately one-third lower.65% similarUnverifiedThe officially claimed 30% reduction in Thinking Token consumption only held true for two of the five problems; for the remaining three, K2.7-Code actually consumed more tokens.59% similarUnverifiedThe model's local throughput is approximately 9 to 22 tokens per second, consumes more tokens than the previous generation, and is slower when thinking mode is enabled.59% similarUnverified端侧推理中吞吐量需达到5-10 token/s对应人类舒适阅读速度，低于此值输出显得一字一顿58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/961194API
curl https://kongchang.com/api/v1/knowledge/claims/961194MCP
get_claim(id=961194)