Unverified50% confidenceFactExact time
When self-written scripts count token usage, if deduplication is not performed, the results are usually higher than the values reported by ccusage or CodeBurn.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
Unverified研究数据显示，在复杂代码生成任务中，将thinking budget从1024 tokens提升到8192 tokens可以带来15-30%的准确率提升，但推理成本增加4-8倍65% similarUnverifiedIn many evaluations conducted on small datasets, scores fluctuate by several percentage points across different runs.64% similarUnverifiedActual token generation speed usually reaches only 50%–70% of the theoretical value, affected by factors such as compute unit scheduling, KV cache reads and writes, and quantization-dequantization overhead.64% similarUnverified推测解码中当草稿通过率低于50%时，实际加速效果会大打折扣甚至产生负优化64% similarUnverifiedIn Claude Code's transcript files, the usage block is repeatedly written in each response chunk line, causing each call to be counted an average of 1.9 times when tallying line by line.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1034950API
curl https://kongchang.com/api/v1/knowledge/claims/1034950MCP
get_claim(id=1034950)