Google DeepMind launches DiffusionGemma, boosting text generation speed by 4x

iconKuCoinFlash
Share
Share IconShare IconShare IconShare IconShare IconShare IconCopy
AI summary iconSummary

expand icon
On-chain news: Google DeepMind has launched DiffusionGemma, an open-source text generation model that increases speed by 4x. Using diffusion technology, it generates 256 tokens in parallel, achieving over 1,000 tokens/s on H100 and 700+ on RTX 5090. The 26B MoE model activates 3.8B parameters during inference and supports consumer GPUs after quantization. It features bidirectional attention and self-correction, making it ideal for code completion and inline editing. The model is open-sourced under Apache 2.0. New token listings may benefit from such advancements in generation efficiency.
ME AI News: Google DeepMind has released the open-source experimental model DiffusionGemma, which employs text diffusion technology to surpass autoregressive token-by-token generation by parallelly generating 256 tokens in a single forward pass. This 26B MoE model activates only 3.8B parameters during inference and, after quantization, fits on consumer-grade GPUs with 18GB VRAM. It achieves over 1000 tokens/s on H100 and over 700 tokens/s on RTX 5090, offering a 4x speed improvement. Featuring bidirectional attention and self-correction capabilities, it is designed for local interactive workflows such as inline editing and code completion, and is released under the Apache 2.0 license. (Source: AiHot)
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.