Gemma
7 articles tagged with this topic
手机本地跑 AI 不再需要联网—— 一个开源安卓应用正在把这件事变得可操作
Pocket LLM v 1.4.0 shrinks to ~200MB, lets users download models on demand and run AI fully offline on Android.
Why some small/medium models fail at grammar checking task?
Gem ma 4B, GPT-OSS-20B, and Qwen3-80B hallucinate spelling errors in grammatically correct sentences.
Gemma 4 'Compliance' Crisis: Fatal Traps in AI Agent Commercialization
Gemma 4's refusal to execute business instructions exposes critical AI agent commercialization risks, forcing enterprises to reassess automation strat
Local LLMs Lose Tool Call Accuracy After 8–9 Chained Calls
Qwen 32B, Gemma 9B, and Command R 32B all fail similarly after 8+ tool calls — attention dilution, not context limits.
OpenCode + Local LLMs: Which Models Work Best for Solo Dev Tasks
A hands-on benchmark of OpenCode with 6+ self-hosted LLMs on an RTX 4080 for real coding tasks.
Real-Time Multimodal AI Runs Locally on M3 Pro with Gemma E2B
Developer runs Gemma 4 E2B locally on Apple M3 Pro for real-time audio/video input with voice output using the Parlor repo.
OpenClaw Runs Local AI Agents on MacBook Air 16GB via TurboQuant
OpenClaw uses llama.cpp TurboQuant cache compression to run agentic AI models on 16GB MacBook Air at 10-15 tokens/sec.