Unverified50% confidenceFactExact time
Qwen3 27B官方模型卡给出了将上下文扩展至100万token的延伸方案
1
Sources
50%
Confidence
Long-term
Relevance
9/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedQwen3.6 27B支持262K上下文长度，34B 80层版本保持相同的上下文长度不变73% similarUnverifiedQwen3-27B原生支持约26万Token的长上下文，通过YaRN等技术可扩展到100万Token量级72% similarUnverifiedQwen-7B经过4-bit量化后，一张搭载24GB显存的消费级显卡即可流畅运行72% similarUnverifiedQwen3.6 34B 80层模型是由社区开发者将Qwen3.6 27B的64层Transformer网络扩展到80层，参数量从27B增长到34B，增幅约26%72% similarUnverifiedQwen3系列编程模型约20GB加载，推理时可膨胀到35-40GB70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/843029API
curl https://kongchang.com/api/v1/knowledge/claims/843029MCP
get_claim(id=843029)