Unverified50% confidenceFactExact time
100万Token意味着模型理论上可一次性处理约50-75万汉字
1
Sources
50%
Confidence
Long-term
Relevance
7/8/2026
First Seen
Sources
GLM-5.2实战指南：100万上下文、额度扣费与切换策略全解析
bilibili开水庆余年6/14/2026
Related Claims
Unverifiedawesome-pretrained-chinese-nlp-models项目已获得超过5500个Star和500+次Fork66% similarUnverified60个Skill/MCP的描述文本总计约占用12,000 tokens（平均每个200 tokens）65% similarUnverified700亿参数的模型按Chinchilla最优对应约1.4万亿训练token65% similarUnverified1M Token约等于75万个英文单词，相当于《哈利·波特》全集的篇幅62% similarUnverified万星导图最高支持10万字的上下文解析62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/213447API
curl https://kongchang.com/api/v1/knowledge/claims/213447MCP
get_claim(id=213447)