Models_And_Releases - AI news 2025-08-05

OpenAI returns to open source roots with new models gpt-oss-120b and gpt-oss-20b

venturebeat.com • Aug 05, 2025

Enterprises can use a powerful, near topline OpenAI LLM on their hardware totally privately and securely, without sending data to the cloud....

TL;DR

OpenAI releases two new open-source large language models, gpt-oss-120b and gpt-oss-20b, with Apache 2.0 licensing, allowing enterprises and developers to use and modify them freely.

Key Takeaways:

These models outperform some of OpenAI's paid models and the competition's offerings, with ranking top performances on reasoning and tool-use benchmarks.
The gpt-oss models are multilingual and perform well across various non-English languages, with potential for fine-tuning to improve regional or linguistic contexts.
OpenAI's shift towards open-source models may signal a reevaluation of its business strategy, as competitors in the open-source AI space continue to gain traction and user adoption.

DeepMind reveals Genie 3, a world model that could be the key to reaching AGI

techcrunch.com • Aug 05, 2025

Google DeepMind has revealed Genie 3, its latest foundation world model that the AI lab says presents a crucial stepping stone on the path to artifici...

TL;DR

Google DeepMind has unveiled Genie 3, a foundation world model that represents a significant step towards artificial general intelligence, enabling the creation of realistic and interactive 3D environments.

Key Takeaways:

Genie 3 can generate real-time, interactive 3D environments at 24 frames per second with a resolution of 720p and multiple minutes of diversity.
The model features 'promptable world events' and can remember what it had previously generated, allowing it to simulate physically consistent environments.
Genie 3's goal is to serve as an ideal training ground for general-purpose agents, pushing them to adapt and learn from their own experience in a realistic way, a crucial step toward achieving artificial general intelligence.

Create personal illustrated storybooks in the Gemini app

blog.google • Aug 05, 2025

Article URL: https://blog.google/products/gemini/storybooks/ Comments URL: https://news.ycombinator.com/item?id=44804411 Points: 166 # Comments: 49...

TL;DR

The Gemini app now offers personalized, illustrated storybooks with read-aloud narration, generated from user descriptions and custom art styles.

Key Takeaways:

Stories can be created in more than 45 languages and various art styles, including pixel art, comics, and claymation.
Users can add a personal touch by uploading their own photos and files for inspiration.
The app is available globally on desktop and mobile in all languages.

Claude Opus 4.1

reddit.com • Aug 05, 2025

TL;DR

NVIDIA announced the Blackwell B200 GPU, promising significant performance increases for large language models and other AI workloads.

Key Takeaways:

Delivers up to 20x improved performance in large language model inference compared to previous architectures.
Features a second-generation Transformer Engine and new tensor core technology for improved efficiency.
Key cloud providers such as AWS, Google Cloud, and Azure have committed to integrating the Blackwell B200 into their cloud offerings.