Updated: 2026-08-13T21:07:57.685Z
Source window: last 36 hours. Summaries used: 3
Google DeepMind had a busy day, launching a new model and shipping a sign language feature for Pixel phones. Meanwhile, Google Research published a study on why large language models sometimes get facts wrong.
- Google DeepMind Launches Gemini 3.7 Flash - Gemini 3.7 Flash is a more intelligent and cost-effective model for coding and agent workflows. It offers improved performance at a lower price point, making it accessible for developers building AI applications.
- Google DeepMind brings sign language AI to Pixel phones - The SL2T model powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, starting with ASL to English. This makes communication more accessible for deaf and hard-of-hearing users.
- Recall, not encoding, limits LLM factuality - Google Research introduces knowledge profiling to show frontier LLMs encode nearly all facts but struggle to recall them, with thinking recovering many failures. This suggests that improving retrieval mechanisms could enhance factual accuracy.
The strongest shared signal is Google's focus on practical AI applications and understanding model limitations. From cost-effective coding models to accessibility features and research on factuality, the company is pushing both performance and reliability.