Unverified50% confidenceFactExact time
MiMo-V2.6模型支持1M（一百万）token的上下文长度
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMiMo-V2.5主模型在双24GB显卡配合96/128GB DDR5内存的VRAM offload配置下，运行速度约为8-10 tokens/s66% similarUnverified目前主流编程模型的上下文窗口从32K到1M Token不等，Qwen系列部分模型支持到1M61% similarUnverifiedMeta的LLaMA 2 70B模型即便使用4-bit量化仍需约35GB显存才能完整加载59% similarUnverifiedSmoLLM是一个LLaMA风格的小型语言模型，参数量为1.09亿（109M）59% similarUnverifiedHy4支持高达100万（1M）token的上下文窗口59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/945492API
curl https://kongchang.com/api/v1/knowledge/claims/945492MCP
get_claim(id=945492)