Expired50% confidenceFactTime unknown
The Gemma 4 1B/2B models include built-in speech recognition and require only 4GB of VRAM after quantization.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Google Gemma 4 Hands-On Review: Offline on Smartphones + Ollama Deployment Tutorial
bilibiliAGI_Ananas
Related Claims
UnverifiedGemma 4采用混合专家架构（MoE），E2B后缀表示Effective 2B，每次推理只激活约20亿有效参数60% similarUnverifiedQ4量化可将模型体积压缩至原来的约1/8，代价是约1-3%的基准测试精度损失59% similarUnverified标准注意力机制的计算复杂度为O(n²)，序列长度翻倍时计算量增加四倍，这是早期GPT-3仅支持4K Token的根本原因59% similarUnverifiedPaliGemma 2 的语言模型部分采用 Gemma 2 系列，参数规模覆盖 3B 至 28B59% similarUnverifiedBlackwell的FP4格式仅有4位（1位符号、2位指数、1位尾数），采用每128个FP4元素共享一个FP8精度缩放因子的块浮点表示58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/57172API
curl https://kongchang.com/api/v1/knowledge/claims/57172MCP
get_claim(id=57172)