Gemma 4 gets a stealth update that fixes tool calling bugs and truncated responses under the same name
Google has quietly updated its Gemma 4 open AI model, improving performance on Nvidia Hopper GPUs and fixing critical bugs related to tool calling and response truncation.
- Gemma 4 has received a stealth update improving Nvidia Hopper GPU performance.
- The update fixes bugs in Gemma 4's tool calling capabilities.
- Truncated response issues in Gemma 4 have also been resolved.
- Google continues to iterate on its open models without explicit announcements.
Google has released an unannounced update to its Gemma 4 open AI model. This update focuses on performance enhancements, particularly for Nvidia Hopper GPUs, aiming to speed up inference.
Crucially, the update addresses significant functional issues. It resolves bugs that affected the model's ability to correctly use tools and fixes problems where responses were being cut short or truncated.
These improvements, delivered without a formal announcement, suggest Google's ongoing commitment to refining its open models for better developer and user experience.
Fixes critical bugs and improves performance for easier integration.
Enhances reliability and efficiency of AI applications using Gemma.
Indicates ongoing development and support for Google's open AI efforts.
Provides a more stable and performant model for learning and experimentation.
Subtle but important improvements to an accessible open AI model.
LLMEx-OpenAI CTO Murati's Thinking Machines drops Inkling, a 975B parameter model that leads US labs but trails China
![[AINews] Thinky's Inkling: 975B-A41B multimodal, new best American Apache 2.0 open model (with Inkling-Small, 276B-A12B)](https://images.weserv.nl/?url=substackcdn.com%2Fimage%2Ffetch%2F%24s_%21AvrX%21%2Cf_auto%2Cq_auto%3Agood%2Cfl_progressive%3Asteep%2Fhttps%253A%252F%252Fsubstack-post-media.s3.amazonaws.com%252Fpublic%252Fimages%252F90048da3-a87f-44d8-8ad4-e954031d2721_2540x1692.png&w=520&fit=cover&q=70&output=webp&dpr=2&we=1&il=1)
LLM[AINews] Thinky's Inkling: 975B-A41B multimodal, new best American Apache 2.0 open model (with Inkling-Small, 276B-A12B)
LLMSoofi Consortium Releases Soofi S 30B-A3B: An Open Hybrid Mamba-Transformer MoE Foundation Model For German And English
Mira Murati’s Thinking Machines draws from Chinese rivals in debut AI model - Financial Times
LLMMeet GPT-Red: an LLM super-hacker OpenAI built to make its models safer
KLAS Global HIT Trends 2026 Report: Artificial Intelligence Becomes the Top Investment Priority - HIT Consultant
The KLAS Global HIT Trends 2026 report identifies artificial intelligence as the leading investment priority for healthcare organizations.
Exclusive: Google DeepMind expands biosecurity effort amid AI safety push - Axios
Google DeepMind is expanding its biosecurity initiatives to address potential biological risks from advanced AI models. This effort is part of a broader internal push for AI safety.
AI ResearchPlease Stop Making Me Opt Out of AI
A Wired article argues that AI features should default to opt-in, not opt-out, to protect user privacy.
ISBE releases guidance on artificial intelligence use in public schools - The Daily Line
The Illinois State Board of Education has released guidance on the use of artificial intelligence in public schools. This guidance aims to help schools navigate the integration of AI in education.
Airbus selects Scaleway to support AI and defence cloud strategy (AIR) - Yahoo Finance UK
Airbus has chosen Scaleway as its cloud provider to support its AI and defence strategy. The partnership aims to enhance Airbus's cloud capabilities.
Microsoft steps up AI rivalry, arms sales teams against OpenAI and Google - ETEnterpriseai.com
Microsoft is equipping its sales teams with advanced AI tools to sharpen its competitive edge against rivals like OpenAI and Google in the rapidly evolving AI market.