Unverified50% confidenceReleaseExact time
项目在Releases中上传了预编译的Windows执行二进制文件和轻量级训练模型权重（3MB和4MB）
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/24/2026
First Seen
Valid until: 11/22/2026
Sources
Related Entities
Related Claims
UnverifiedQwen3.5-9B contains approximately 9 billion trainable parameters, and after 4-bit quantization, the model file size can be compressed to approximately 5–6 GB.71% similarUnverifiedCombined with interpolation techniques such as YaRN and Dynamic NTK, models trained with a context length of 4K can be extended to context windows of 128K or even 1M.61% similarUnverifiedUnsloth 通过 4-bit/8-bit 量化感知训练与动态激活卸载技术，可将 7B 模型的微调显存需求压缩至约 4–6 GB61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/795575API
curl https://kongchang.com/api/v1/knowledge/claims/795575MCP
get_claim(id=795575)