Unverified50% confidenceFactExact time
128K指模型的上下文窗口大小，即模型单次可以处理约128,000个Token的输入信息
1
Sources
50%
Confidence
Long-term
Relevance
7/17/2026
First Seen
Sources
Related Claims
UnverifiedWhen models with hundreds of billions of parameters process a 128K context window, KV Cache can occupy tens of gigabytes of VRAM.81% similarUnverifiedThe latest large models have context windows of 128K or even 200K tokens, yet still face attention dilution issues when handling complex engineering tasks.76% similarVerifiedThe context windows of current mainstream large language models have expanded to 128K or even 200K tokens.76% similarVerifiedMainstream large models already support context windows of 128K or even longer.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/538941API
curl https://kongchang.com/api/v1/knowledge/claims/538941MCP
get_claim(id=538941)