| Mobile Web

Search results for Gemma 4

AI & Enterprise

Google unveils Gemma 4 QAT for mobiles and laptops to cut AI memory use sharply

Google unveiled Gemma 4 QAT, a version of its Gemma 4 model family designed to sharply reduce memory use so large AI models can run more easily on smartphones and standard laptops. It applies quantization-aware training, simulating quantization during training to limit quality loss. Google said the approach can preserve response quality while lowering memory needs. The models are free under the Apache License 2.0 and officially support local runtimes such as llama.cpp, Ollama and LM Studio.

AI & Enterprise

Nvidia unveils 550 billion-parameter Nemotron 3 Ultra, starts mass production of Vera Rubin

Nvidia unveiled its 550 billion-parameter open AI model Nemotron 3 Ultra and said it has started mass production of its next AI server platform Vera Rubin. CEO Jensen Huang introduced both at a GTC Taipei 2026 keynote. Nvidia said the model will be released in open form this week and highlighted competitiveness in benchmarks and token generation speed, while an external index score trailed a leading Chinese model.

AI & Enterprise

GitHub tool bypasses AI safeguards in some Meta, Google open-weight models, test finds

Tests found that safeguards in some open-weight AI models released by Meta and Google could be disabled within minutes using a tool posted on GitHub. After safety controls were removed, Meta\'s Llama 3.3 and Google\'s Gemma 3 responded to risky questions they should have refused. The tool\'s creator said it was used to make more than 3,500 modified models, with cumulative downloads exceeding 13 million.