DeepMind unveils DiffusionGemma and demos Genie-3 worlds
nathanbenaich · x · 2026-07-20
At RAAIS 2026, Raia Hadsell announced that DeepMind had shipped a brand-new model less than 48 hours earlier: DiffusionGemma, a 26B text-diffusion model that can correct its own mistakes during generation.
She also mentioned Genie-3 worlds grounded on Street View, including a demo of diving under the Golden Gate Bridge, and argued that an autoregressive model can never solve Sudoku.
More from Models
- Claim says Kimi was distilled from Fable, sparking a model-attribution jab — cephaloform · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Macaron V1 adds LoRA RL on GLM 5.2 and claims SOTA benchmark gains — Xianbao_QIAN · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22