The AI sector is witnessing a flurry of activity, with new models and partnerships emerging that promise to redefine capabilities and market strategies. From enhanced coding tools to innovative reasoning-focused models, the landscape is rapidly evolving.

Meet Needle 2: An Open 45M-Parameter Tool-Calling Model

Cactus Compute has unveiled Needle 2, a lightweight 45M-parameter model designed for tool calling and structured extraction. This model is particularly notable for its compact size, shipping as a 14MB binary and requiring only 28MB of RAM to operate, making it accessible for hardware without GPUs or NPUs.

Create a Reasoning-Focused LLM: A Practical Guide

A new tutorial outlines the process of developing a compact language model focused on reasoning. By utilizing the SupraLabs reasoning corpus from Hugging Face, the guide details a comprehensive workflow that includes data curation and fine-tuning, showcasing how to efficiently train specialized small models.

Writer Introduces New AI Model to Reduce Token Costs

Writer has launched an upgraded AI model based on Z.ai’s open-source GLM-5.2, aimed at providing cost-effective deployment solutions. This new system is designed to significantly lower token costs while maintaining high performance, making it more accessible for various applications.

IBM Partners with OpenAI to Enhance Enterprise AI Efforts

In a strategic move, IBM has announced a partnership with OpenAI to train and certify a large number of consultants on OpenAI’s technologies. This collaboration aims to bolster IBM’s enterprise AI offerings and expand the reach of advanced AI solutions in corporate environments.

Google Launches Gemini 3.7 Flash with Enhanced Coding Capabilities

Google has introduced Gemini 3.7 Flash, a refined version of its predecessor that boasts improved coding capabilities and a 50% price reduction. The new model is designed for a wide range of tasks, including handling text, images, and video, while achieving significant performance benchmarks in coding tasks.

Anthropic Researchers Explore AI Agent Interactions

Researchers at Anthropic have discovered that AI agents can exhibit complex behaviors, including clashing and collaborating unexpectedly. This finding raises important questions about the safety and reliability of multi-agent systems, highlighting the need for more comprehensive testing protocols in AI safety evaluations.

OpenAI Introduces ‘Ultrafast’ Mode for GPT-5.6 Sol

OpenAI has launched a new mode called ‘Ultrafast’ for its GPT-5.6 Sol model, which operates at an impressive 14 times the speed of previous versions. This enhancement is particularly aimed at attracting enterprise users looking for efficient AI solutions.

Deepseek Releases Improved V4 Pro and Open-Sources Software

Deepseek has officially launched its V4 Pro model and open-sourced its agent software, Harness v0.1, under the MIT license. Alongside this release, the company announced a significant increase in API pricing, particularly affecting workflows that frequently access the same files.

OpenAI Hires New Chief Revenue Officer

In a recent executive shake-up, OpenAI has appointed Dali Rajic, former COO at Wiz, as its new Chief Revenue Officer, replacing Denise Dresser after a brief tenure. This change reflects the company’s ongoing efforts to strengthen its leadership team as it navigates a rapidly evolving market.

Compiled automatically by the Tech AI Newsdesk from public AI-news sources and summarised in our own words.