Quanta AI Labs

Google · Text Generation

DiffusionGemma

Experimental Apache-2.0 diffusion-based text-generation model (26B MoE, 3.8B active, ~18GB VRAM quantized) generating 256-token blocks in parallel via iterative refinement with bidirectional attention — 1000+ tok/s on H100, ~700 on RTX 5090 (up to 4x faster than autoregressive). Built on Gemma 4 + Gemini Diffusion research; excels at speed-critical infilling/inline edits.

Open the source ↗

Similar Language Models releases

  • SenseNova-Loop-9B-Preview

    OpenSenseNova (SenseNova) · Looped multimodal LLM · Week 45, 2026

  • Spark-X2.6-1.7B

    XHToken · On-device LLM · Week 45, 2026

  • LFM3

    Liquid AI · Small multimodal LFMs · Week 45, 2026

  • Nanbeige4.3-33B

    Nanbeige Team (Kanzhun) · Agentic 33B LLM · Week 45, 2026

  • Gemini 4 Carbon

    Google DeepMind · Frontier cyber LLM · Week 45, 2026

  • Carbon-X

    Hugging Face (HuggingFaceBio) · DNA Language Model · Week 44, 2026

More from Google

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.