Models
After dissing Anthropic for limiting Mythos, OpenAI restricts access to Cyber, too
OpenAI is restricting access to its Cyber model after previously criticizing Anthropic for limiting Mythos. This move signals a tightening of access to AI models, impacting how developers and researchers can utilize these tools.
Elon Musk testifies that xAI trained Grok on OpenAI models
Elon Musk testifies that xAI trained its Grok chatbot using OpenAI models. This means Grok leverages established AI capabilities to enhance its performance and user interactions.
Where the goblins came from
OpenAI is tightening Codex's outputs to prevent it from generating irrelevant content like random goblin references. This change aims to improve usability in real workflows for developers.
OpenAI Really Wants Codex to Shut Up About Goblins
OpenAI is addressing concerns regarding its Codex model, particularly its tendency to generate inappropriate or irrelevant content, such as references to goblins. The company is focused on refining Codex to improve its performance and ensure it aligns more closely with user expectations and needs.
Here is what an LLM that knows nothing after 1930 thinks our world looks like in 2026
The article explores the perspective of a large language model (LLM) that has no knowledge of events or developments after 1930, speculating on what it might envision for the world in 2026. It highlights the limitations of such an LLM in understanding contemporary society and technological advancements. The discussion emphasizes the gap between historical knowledge and current realities in AI models.
NVIDIA Nemotron 3 Nano Omni model now available on Amazon SageMaker JumpStart
The NVIDIA Nemotron 3 Nano Omni model has been released and is now available for use on Amazon SageMaker JumpStart. This new model aims to enhance machine learning capabilities for developers and researchers by providing an accessible platform for deploying AI solutions.
Introducing GPT-5.5
OpenAI has announced the release of GPT-5.5, an advanced version of its language model that offers improved performance and capabilities over its predecessors. This update is expected to enhance various applications, making AI interactions more seamless and effective for users.
GPT 5.5
The article discusses the anticipated release of GPT 5.5, highlighting its expected improvements in natural language understanding and generation capabilities. It also explores potential applications and implications for various industries, emphasizing the ongoing evolution of AI technologies. The release is expected to further enhance user experience and productivity across different sectors.
[AINews] OpenAI launches GPT-Image-2
OpenAI has officially launched GPT-Image-2, an advanced AI model designed for generating and manipulating images based on textual descriptions. This new model aims to enhance creative workflows and provide users with more powerful tools for image generation.
GPT-Image-2
GPT-Image-2 is a new AI model that enhances image generation capabilities, allowing for more detailed and contextually relevant outputs. This update aims to improve user experience and expand the potential applications of AI in creative fields.
[AINews] Moonshot Kimi K2.6: the world's leading Open Model refreshes to catch up to Opus 4.6 (ahead of DeepSeek v4?)
Moonshot has released an updated version of its leading Open Model, Kimi K2.6, aiming to compete with Opus 4.6 and potentially ahead of DeepSeek v4. This refresh highlights the ongoing advancements in AI model development and the competitive landscape among leading AI frameworks.
[AINews] Anthropic Claude Opus 4.7 - literally one step better than 4.6 in every dimension
Anthropic has released Claude Opus 4.7, which boasts improvements over its predecessor, 4.6, in every dimension. This update highlights the company's commitment to enhancing their AI model's capabilities.
Anthropic's Claude Opus 4.7
Anthropic has released Claude Opus 4.7, an updated version of its AI model that includes enhancements in performance and capabilities. This update aims to improve user experience and expand the model's applications across various domains.
[AINews] Top Local Models List - April 2026
The article presents a list of the top local AI models as of April 2026, highlighting their performance and unique features. It emphasizes the growing importance of localized AI solutions in various sectors and their impact on regional industries.
Gemini Robotics-ER 1.6: Powering real-world robotics tasks through enhanced embodied reasoning
Gemini Robotics-ER 1.6 introduces advanced embodied reasoning capabilities that enhance the performance of robotics in real-world tasks. This update aims to improve the interaction and adaptability of robots in various environments, showcasing significant advancements in AI-driven robotics technology.
Multimodal Embedding & Reranker Models with Sentence Transformers
Hugging Face has introduced new multimodal embedding and reranker models using Sentence Transformers, enhancing the capability to process and rank information from various data types. These advancements aim to improve the performance of AI applications in understanding and organizing content across different modalities.
[AINews] Meta Superintelligence Labs announces Muse Spark, first frontier model on their completely new stack
Meta Superintelligence Labs has unveiled Muse Spark, the first frontier model built on their new technological stack. This development marks a significant step in advancing AI capabilities and showcases Meta's commitment to innovation in the field.
Google announces Gemma 4 open AI models, switches to Apache 2.0 license
Google has announced the release of its Gemma 4 open AI models and has transitioned to using the Apache 2.0 license for these models. This move is expected to enhance collaboration and accessibility in the AI community.
Gemma 4: Byte for byte, the most capable open models
Google DeepMind has announced Gemma 4, claiming it to be the most capable open model to date, outperforming previous iterations in various benchmarks. This new model is designed to enhance the capabilities of AI applications and improve user interactions.
Welcome Gemma 4: Frontier multimodal intelligence on device
Hugging Face has introduced Gemma 4, a new multimodal intelligence model designed to operate directly on devices. This advancement aims to enhance user experience by enabling more efficient and responsive AI interactions without relying on cloud processing.
Falcon Perception
Hugging Face has introduced Falcon Perception, a new AI model designed to enhance visual understanding in various applications. This model aims to improve the capabilities of existing AI systems in processing and interpreting visual data more effectively.
Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents
Granite 4.0 3B Vision is a new multimodal AI model designed to enhance the processing of enterprise documents by integrating visual and textual data. This model aims to improve efficiency and accuracy in document management tasks across various industries.
Gemini 3.1 Flash-Lite: Built for intelligence at scale
Google DeepMind has announced the release of Gemini 3.1 Flash-Lite, a new AI model designed to enhance intelligence at scale. This update focuses on improving efficiency and performance in various applications, making it a significant advancement in the Gemini series.
Nano Banana 2: Combining Pro capabilities with lightning-fast speed
Google DeepMind has announced the release of Nano Banana 2, an advanced AI model that integrates professional capabilities with enhanced speed. This new model aims to improve efficiency and performance in various applications, showcasing the company's commitment to innovation in AI technology.
Gemini 3.1 Pro: A smarter model for your most complex tasks
Google DeepMind has introduced Gemini 3.1 Pro, an advanced AI model designed to handle complex tasks more effectively. This update aims to enhance performance and provide users with smarter solutions for challenging applications.
Gemini 3 Deep Think: Advancing science, research and engineering
Google DeepMind has introduced Gemini 3 Deep Think, an advanced AI model designed to enhance scientific research and engineering capabilities. This model aims to support researchers by providing sophisticated analytical tools and insights, potentially accelerating discoveries across various fields.
Gemini 3 Flash: frontier intelligence built for speed
Google DeepMind has introduced Gemini 3 Flash, an advanced AI model designed to enhance speed and efficiency in processing information. This new model aims to push the boundaries of AI capabilities, focusing on rapid intelligence deployment across various applications.
Reaching Across the Isles: UK-LLM Brings AI to UK Languages With NVIDIA Nemotron
NVIDIA has introduced the UK-LLM, a new AI model designed to enhance language processing for various UK languages. This initiative aims to improve accessibility and usability of AI technologies across different linguistic communities in the UK.
NVIDIA Releases New AI Models and Developer Tools to Advance Autonomous Vehicle Ecosystem
NVIDIA has unveiled new AI models and developer tools aimed at enhancing the autonomous vehicle ecosystem. These advancements are designed to support developers in creating more sophisticated and efficient autonomous driving systems.