Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders.<\/em> Subscribe Now<\/em><\/p>\n\n\n\n Google\u2019s DeepMind AI research team<\/strong> has unveiled a new open source AI model today, Gemma 3 270M.<\/strong><\/p>\n\n\n\n As its name would suggest, this is a 270-million-parameter model<\/strong> \u2014 far smaller than the 70 billion or more parameters of many frontier LLMs (parameters being the number of internal settings governing the model\u2019s behavior).<\/p>\n\n\n\n While more parameters generally translates to a larger and more powerful model, Google\u2019s focus with this is nearly the opposite: high-efficiency, giving developers a model small enough to run directly on smartphones<\/strong> and locally<\/strong>, without an internet connection<\/strong>, as shown in internal tests on a Pixel 9 Pro SoC.<\/p>\n\n\n\n Yet, the model is still capable of handling complex, domain-specific tasks and can be quickly fine-tuned in mere minutes to fit an enterprise or indie developer\u2019s needs.<\/p>\n\n\n\n Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:<\/p>\n\n\n\n Secure your spot to stay ahead<\/strong>: https:\/\/bit.ly\/4mwGngO<\/p>\n\n\n\n On the social network X, Google DeepMind Staff AI Developer Relations Engineer Omar Sanseviero added that it Gemma 3 270M can also run directly in a user\u2019s web browser, on a Raspberry Pi<\/strong>, and \u201cin your toaster,\u201d underscoring its ability to operate on very lightweight hardware.<\/p>\n\n\n\n Gemma 3 270M combines 170 million embedding parameters \u2014 thanks to a large 256k vocabulary capable of handling rare and specific tokens \u2014 with 100 million transformer block parameters. <\/p>\n\n\n\n According to Google, the architecture supports strong performance on instruction-following tasks right out of the box while staying small enough for rapid fine-tuning and deployment on devices with limited resources, including mobile hardware.<\/p>\n\n\n\n Gemma 3 270M inherits the architecture and pretraining of the larger Gemma 3 models, ensuring compatibility across the Gemma ecosystem. With documentation, fine-tuning recipes, and deployment guides available for tools like Hugging Face, UnSloth, and JAX, developers can move from experimentation to deployment quickly.<\/p>\n\n\n\n The score places it well above similarly small models like SmolLM2 135M Instruct and Qwen 2.5 0.5B Instruct<\/strong>, and closer to the performance range of some billion-parameter models, according to Google\u2019s published comparison.<\/p>\n\n\n\n However, as researchers and leaders at rival AI startup Liquid AI pointed out in replies on X, Google left off Liquid\u2019s own LFM2-350M model released back in July of this year, which scored a whopping 65.12%<\/strong> with just a few more parameters (similar sized language model, however).<\/p>\n\n\n\n One of the model\u2019s defining strengths is its energy efficiency. In internal tests using the INT4-quantized model on a Pixel 9 Pro SoC, 25 conversations consumed just 0.75% of the device\u2019s battery. <\/strong><\/p>\n\n\n\n This makes Gemma 3 270M a practical choice for on-device AI, particularly in cases where privacy and offline functionality are important.<\/p>\n\n\n\n The release includes both a pretrained and an instruction-tuned model, giving developers immediate utility for general instruction-following tasks. <\/p>\n\n\n\n Quantization-Aware Trained (QAT) checkpoints are also available, enabling INT4 precision with minimal performance loss and making the model production-ready for resource-constrained environments.<\/p>\n\n\n\n Google frames Gemma 3 270M as part of a broader philosophy of choosing the right tool for the job rather than relying on raw model size. <\/p>\n\n\n\n For functions like sentiment analysis, entity extraction, query routing, structured text generation, compliance checks, and creative writing, the company says a fine-tuned small model can deliver faster, more cost-effective results than a large general-purpose one.<\/p>\n\n\n\n The benefits of specialization are evident in past work, such as Adaptive ML\u2019s collaboration with SK Telecom. <\/p>\n\n\n\n By fine-tuning a Gemma 3 4B model for multilingual content moderation, the team outperformed much larger proprietary systems. <\/p>\n\n\n\n Gemma 3 270M is designed to enable similar success at an even smaller scale,<\/strong> supporting fleets of specialized models tailored to individual tasks.<\/p>\n\n\n\n Beyond enterprise use, the model also fits creative scenarios. In a demo video posted on YouTube, Google shows off a Bedtime Story Generator app built with Gemma 3 270M and Transformers.js that runs entirely offline in a web browser,<\/strong> showing the versatility of the model in lightweight, accessible applications.<\/p>\n\n\n\n \n
\n<\/div>
\n\n\n\n\n
\n<\/div>High scores on benchmarks for its size, and high hefficiency<\/h2>\n\n\n\n
A small, fine-tuned version of Gemma 3 270M can perform many functions of larger LLMs<\/h2>\n\n\n\n
Demo Bedtime Story Generator app shows off the potential of Gemma 3 270M<\/h2>\n\n\n\n