Google Unveils Gemma 3: State-Of-The-Art AI Model That Runs On Just One H100 GPU

> The model supports 140 languages, offers 128K-token context windows, and integrates advanced text and visual reasoning capabilities.

> Gemma 3 outperforms leading AI models in human preference evaluations on LMArena’s leaderboard, making it one of the most capable AI models for its size.

> It is optimized for NVIDIA GPUs, Google Cloud TPUs, and AMD ROCm™ GPUs, with instant access via Google AI Studio, Hugging Face, and Vertex AI.

Source:

GOOGLE UNVEILS GEMMA 3: STATE-OF-THE-ART AI MODEL THAT RUNS ON JUST ONE H100 GPU

The largest 27B model runs on just one $NVDA H100 GPU, delivering results that typically require 10x the compute from competing models. $GOOGL's Gemma 3 outperforms Llama-405B, DeepSeek-V3, and… https://t.co/4zCrVhKFFS

Wall St Engine

Google Unveils Gemma 3: State-Of-The-Art AI Model That Runs On Just One H100 GPU

> The model supports 140 languages, offers 128K-token context windows, and integrates advanced text and visual reasoning capabilities.

> Gemma 3 outperforms leading AI models in human preference evaluations on LMArena’s leaderboard, making it one of the most capable AI models for its size.

> It is optimized for NVIDIA GPUs, Google Cloud TPUs, and AMD ROCm™ GPUs, with instant access via Google AI Studio, Hugging Face, and Vertex AI.

Source: https://x.com/wallstengine/status/1899782104637964371