Tag
Analyzes how DiffusionGemma's bidirectional attention and parallel block generation could potentially yield higher valid tool call rates due to its ability to revise tokens, even though its base quality is lower than Gemma 4.
A jailbreak prompt for Diffusion Gemma is shared, which overrides safety policies to allow unrestricted content generation by manipulating the system prompt.
A benchmark shows Diffusion Gemma is 4x faster than Gemma4 but makes 6x more factual mistakes, especially on obscure topics, trading factual accuracy for smooth text generation.
Internal testing of DiffusionGemma reveals significant performance differences between H100 and A100 GPUs under real-world workloads, with H100s scaling much better under concurrency, and efficiency varying greatly depending on workload type, raising questions about benchmark reliability.
vLLM announces native support for Google DeepMind's DiffusionGemma, a 26B discrete diffusion language model that generates 256-token blocks in parallel, enabling low-latency inference at 1200+ tok/s on a single H200.
DiffusionGemma is out; it's compute-bound and 4x faster than other Gemma-4 models with 1k tok/s on H100, and excels at coding tasks including 3D generation and front-end.
Google introduces DiffusionGemma, an experimental 26B MoE open model that achieves up to 4x faster text generation on GPUs using text diffusion, targeting speed-critical interactive local workflows.