AI News

Follow the latest AI and automation news

8 Jul 202623 Muharram 1448 AH

ChatGPT’s upgraded voice mode is better at shutting up

OpenAI has announced an update to ChatGPT's voice mode with the introduction of the GPT-Live-1 model, designed to reduce interruptions during conversations for smoother interactions. GPT-Live-1 can wait for users to finish speaking, creating a more natural conversational experience. During a press conference, Kundan Kumar, Head of Research at OpenAI, described it as "the smartest voice model" to date. The new model will automatically convert queries into the best text models, such as GPT-5.5, when needed, allowing for swift transitions from research to discussing results, reflecting a commitment to enhancing human-AI interaction.

Models

8 Jul 202623 Muharram 1448 AH

Anthropic's fix for Fable 5's high cost is turning it into a manager that delegates to Sonnet 5

Anthropic has recommended using Claude Fable 5 primarily as a manager for smaller models rather than deploying it for every task. This approach combines Fable 5 with Sonnet 5 in an "Advisor" pattern, achieving 92% of Fable 5's performance at only 63% of the cost. This shift in strategy addresses the high costs associated with Fable 5, potentially influencing deployment strategies in the AI market.

Agents

8 Jul 202623 Muharram 1448 AH

Prime Intellect raises $130M Series A to help enterprises build their own AI agents

Founded in 2024, Prime Intellect aims to empower organizations to train their own agentic systems without reliance on advanced AI labs. The company seeks to provide tools and resources that enable organizations to build intelligent systems capable of performing multiple tasks independently, enhancing their competitive edge. This development marks a significant step towards enabling institutions to leverage AI directly, potentially improving efficiency and reducing costs associated with collaboration with major AI labs.

💡

Flint: A visualization language for the AI era

How might Flint change the future of chart creation?

Business

Anthropic's Claude Fable 5 dominates new industry benchmarks at a steep premium

8 Jul 202623 Muharram 1448 AH

Business

8 Jul 202623 Muharram 1448 AH

Anthropic's Claude Fable 5 dominates new industry benchmarks at a steep premium

Anthropic's Claude Fable 5 has topped all six new industry-specific performance indices released by Artificial Analysis, covering finance, law, and medicine. However, this leadership comes at a steep price, with a single task in the Strategy & Ops Index costing $3.48, over a hundred times the $0.03 charged by DeepSeek V4 Pro. Despite the high cost, the score difference is only 12 points, raising questions about the value of investing in Claude Fable 5 compared to more affordable options.

Models

NVIDIA Nemotron Achieves Benchmark-Leading Performance With LangChain Deep Agents Harness

8 Jul 202623 Muharram 1448 AH

Models

8 Jul 202623 Muharram 1448 AH

NVIDIA Nemotron Achieves Benchmark-Leading Performance With LangChain Deep Agents Harness

NVIDIA has announced the Nemotron 3 Ultra model, which delivers benchmark-leading performance at lower costs compared to closed models. The Deep Agents system from LangChain has been optimized to work with Nemotron 3 Ultra, achieving the highest accuracy among open models. It also completes more tasks at a tenfold lower inference cost than leading closed models. The LangChain platform enables over 200 million monthly users to access high-performance models, allowing teams to conduct continuous evaluations and faster experimentation.

Agents

Google Deepmind adds background execution and MCP support to Gemini API managed agents

8 Jul 202623 Muharram 1448 AH

Agents

8 Jul 202623 Muharram 1448 AH

Google Deepmind adds background execution and MCP support to Gemini API managed agents

Google DeepMind has announced the addition of four new features to its Gemini API managed agents. These features include asynchronous background execution, direct connection to remote MCP servers, the ability to use custom functions with sandbox tools, and credential refresh without losing state. These updates significantly enhance agent performance, enabling more efficient operation and improved user experience. They also facilitate the development of AI-driven applications, making them more versatile.

Agents

Anthropic Launches Claude Cowork for Enhanced Task Management

8 Jul 202623 Muharram 1448 AH

Agents

8 Jul 202623 Muharram 1448 AH

Anthropic Launches Claude Cowork for Enhanced Task Management

Anthropic has launched its intelligent personal assistant, Claude Cowork, on the web and smartphones, expanding its availability beyond just desktop applications. This move aims to facilitate task management from anywhere, reflecting a shift towards enhancing user experience. Claude Cowork can perform multi-step tasks such as organizing files, researching topics, and preparing reports, thereby increasing work efficiency. This development comes at a time when reliance on technology in modern workplaces is growing.

Governance

Our approach to government and national security partnerships

8 Jul 202623 Muharram 1448 AH

Governance

8 Jul 202623 Muharram 1448 AH

Our approach to government and national security partnerships

OpenAI outlines its approach to partnerships with governments, emphasizing principles of responsible AI use. These principles include democratic accountability and public safety, reflecting the company's commitment to transparency. The organization aims to forge strong relationships with government entities by developing policies that support ethical technology use. This includes collaborative projects designed to enhance national security and improve crisis response.

Models

Separating signal from noise in coding evaluations

8 Jul 202623 Muharram 1448 AH

Models

8 Jul 202623 Muharram 1448 AH

Separating signal from noise in coding evaluations

A new study from OpenAI reveals reliability and accuracy issues with SWE-Bench Pro, a widely used benchmark for evaluating AI models. The analysis identifies problems such as the inability to accurately measure performance, raising concerns about the effectiveness of this benchmark in assessing programming models. Given its importance in the AI community, widely utilized by researchers and developers, these findings prompt a reevaluation of SWE-Bench Pro's role as a primary performance evaluation standard, potentially impacting strategic decisions in software development.