Unverified60% confidenceFactExact time
After Gemma 4 QAT optimization, the minimum memory footprint can be reduced to 1GB
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
GPT 5.6 Internal Testing Codename Revealed, Google Pays SpaceX $920M Monthly for Computing Power
bilibiliinfinite灵感港6/6/2026
Related Claims
UnverifiedThrough targeted two-bit compression techniques, the Gemma 4 E-to-B model's memory footprint has been compressed to approximately 1GB79% similarUnverified模型量化技术将 Gemma 2B 的内存占用从约 5GB 降至 1.5GB 左右79% similarVerified4-bit量化使模型存储空间缩减为FP16的约1/472% similarPartially Verified标准的FP32模型每个参数占用4字节，而INT4量化将其压缩至0.5字节，理论上可实现8倍的内存节省72% similarVerifiedQ4_K_M 采用混合精度，部分关键权重以 6bit 存储，其余以 4bit 存储，整体平均约 4.5bit71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/50846API
curl https://kongchang.com/api/v1/knowledge/claims/50846MCP
get_claim(id=50846)