ended5월 20일· 1 sources
Gemma 4: The Dawn of Truly Reliable Open-Weight AI Agents
Gemma 4, 단순 LLM을 넘어 '자율 에이전트'의 기준을 바꾸다
Why it matters
Gemma 4 represents a paradigm shift from text generation to autonomous task execution, evidenced by its massive leap in τ2-bench scores. By embedding native function calling and long-form reasoning into the architecture, Google has democratized the ability to build production-grade AI agents that were previously exclusive to closed-frontier models.
1
Sources
+0
24h
—
Growth
124d
Active
Gemma 4τ2-benchAgentic tool useMoE architectureGoogle DeepMindFunction calling