https://yakinin.com/en/tags/%23edgeai/ Recent content in #EdgeAI on Home Hugo -- 0.148.2 en Fri, 05 Sep 2025 19:54:00 +0000 https://yakinin.com/en/posts/20250905-google-embeddinggemma-on-device-rag/ Fri, 05 Sep 2025 19:54:00 +0000 https://yakinin.com/en/posts/20250905-google-embeddinggemma-on-device-rag/ <p>I usually default to OpenAI for embeddings, but Google&rsquo;s new EmbeddingGemma model is a noteworthy development. It’s not just another model; it’s a strategic move that shows real promise for improving Retrieval-Augmented Generation (RAG) pipelines, especially in on-device and edge applications.</p> <h2 id="what-is-embeddinggemma">What is EmbeddingGemma?</h2> <p>Google has released EmbeddingGemma as a lightweight, efficient, and multilingual embedding model. At just 308M parameters, it’s designed for high performance in resource-constrained environments. This isn&rsquo;t just about making a smaller model; it&rsquo;s about making a <em>capable</em> small model.</p> https://yakinin.com/en/posts/20250801-deepseek-v3-local-performance/ Thu, 27 Mar 2025 11:22:11 +0000 https://yakinin.com/en/posts/20250801-deepseek-v3-local-performance/ <div style="display: flex; justify-content: center; gap: 1em; flex-wrap: wrap;"> <img src="https://yakinin.com/img/20250801-deepseek-v3-local-performance-0.jpg" style="max-width: 350px; width: 100%;" /> </div> <p>DeepSeek has once again followed its &ldquo;quiet release&rdquo; strategy, making its new DeepSeek-V3-0324 model available on Hugging Face without any major announcements. Instead of marketing hype, they&rsquo;ve simply delivered a solution for the community to evaluate.</p> <p>I tested the model locally on a Mac Studio equipped with an M3 Ultra chip and saw impressive performance, generating over 20 tokens per second. This marks a significant acceleration for running capable models on local hardware, making it a viable option for developers.</p>