All posts tagged: Qwen

Deepseek v4 Performance Analysis: Does It Beat Kimi K2.6 and Qwen 3.6 Plus?

Deepseek v4 Performance Analysis: Does It Beat Kimi K2.6 and Qwen 3.6 Plus?

Published by skeptic

Deepseek v4 has officially undergone comprehensive testing, revealing both its potential and its limitations. Developed as an open source AI model, it is available in two versions: the high-performance Deepseek v4 Pro and the cost-efficient Deepseek v4 Flash. The Pro model, with its 1.6 trillion parameters and focus on advanced tasks like STEM applications and code generation, aims to cater to demanding use cases. Meanwhile, the Flash model offers a streamlined alternative with 284 billion parameters, targeting users with simpler needs. However, as highlighted by World of AI, real-world testing has exposed critical gaps in performance, particularly in areas requiring creativity, nuanced reasoning, or precision. Explore the strengths and weaknesses of Deepseek v4 through a closer look at its pricing structure, task-specific performance and how it compares to competitors like Kimi K2.6 and Opus 4.6. Gain insight into why the Pro model struggles with consistency despite its technical specifications and learn how the Flash model balances affordability with practical constraints. This breakdown also examines where Deepseek v4 excels, such as long-context processing and considers what …

Qwen 3.6 Plus : 1M Context Window & Agentic Coding Tools

Qwen 3.6 Plus : 1M Context Window & Agentic Coding Tools

Published by skeptic

Qwen 3.6 Plus has arrived, bringing a host of updates tailored for developers and researchers tackling technical challenges. This latest iteration, as highlighted by Prompt Engineering, emphasizes structured problem-solving through features like agentic coding, which facilitates step-by-step refinement for intricate workflows. With its 1-million-token context window, the model can handle extensive datasets while maintaining coherence, making it particularly effective for tasks such as simulations and large-scale data analysis. While not designed for conversational AI, its specialized focus positions it as a reliable choice for users requiring precision and advanced reasoning. In this feature, you’ll explore how Qwen 3.6 Plus excels in multimodal understanding, integrating text, images and videos to address diverse technical and creative challenges. Gain insight into its real-world applications, from real-time tracking of the International Space Station to generating detailed datasets for creative projects. Additionally, the discussion will touch on its limitations, such as its dependency on specific frameworks, making sure a balanced understanding of its capabilities. This breakdown offers a comprehensive look at how Qwen 3.6 Plus can support your most demanding …

Google Gemma 4, Anthropic’s Secret Al Agent, Qwen 3.6 & More

Google Gemma 4, Anthropic’s Secret Al Agent, Qwen 3.6 & More

Published by skeptic

Artificial intelligence continues to evolve rapidly, with recent developments showcasing significant progress across multimodal models, persistent agents and advanced coding workflows. Universe of AI explores key innovations, including Google’s Gemma 4, a multimodal AI model optimized for diverse inputs like audio, video and images. Notably, Gemma 4 combines efficiency with accessibility, running effectively on consumer hardware while offering features like extended context windows and native function calling. This balance of performance and usability positions it as a noteworthy step forward in making AI more practical for everyday applications. Dive into this explainer to gain insight into how Anthropic’s persistent AI agent, Conway, introduces always-on functionality for real-time responsiveness and how Alibaba’s Qwen 3.6 Plus uses agentic coding to streamline complex development workflows. You’ll also discover Z.AI’s GLM 5V Turbo, which integrates vision-to-code capabilities to bridge the gap between design and implementation. These advancements highlight the diverse ways AI is reshaping automation, engineering and productivity, offering a detailed look at the technologies driving the next wave of innovation. Google’s Gemma 4: A Multimodal Marvel TL;DR Key …

Nvidia’s new open weights Nemotron 3 super combines three different architectures to beat gpt-oss and Qwen in throughput

Nvidia’s new open weights Nemotron 3 super combines three different architectures to beat gpt-oss and Qwen in throughput

Published by skeptic

Multi-agent systems, designed to handle long-horizon tasks like software engineering or cybersecurity triaging, can generate up to 15 times the token volume of standard chats — threatening their cost-effectiveness in handling enterprise tasks. But today, Nvidia sought to help solve this problem with the release of Nemotron 3 Super, a 120-billion-parameter hybrid model, with weights posted on Hugging Face. By merging disparate architectural philosophies—state-space models, transformers, and a novel “Latent” mixture-of-experts design—Nvidia is attempting to provide the specialized depth required for agentic workflows without the bloat typical of dense reasoning models, and all available for commercial usage under mostly open weights. Triple hybrid architecture At the core of Nemotron 3 Super is a sophisticated architectural triad that balances memory efficiency with precision reasoning. The model utilizes a Hybrid Mamba-Transformer backbone, which interleaves Mamba-2 layers with strategic Transformer attention layers. To understand the implications for enterprise production, consider the “needle in a haystack” problem. Mamba-2 layers act like a “fast-travel” highway system, handling the vast majority of sequence processing with linear-time complexity. This allows the model …

Watch Out, Meta. I Tried Alibaba’s Qwen Smart Glasses and They’re Mega Impressive

Watch Out, Meta. I Tried Alibaba’s Qwen Smart Glasses and They’re Mega Impressive

Published by skeptic

Mobile World Congress in Barcelona might be a European tech show, but for the past few years, the event has largely been dominated by Chinese phone companies such as Xiaomi and Honor. This year, they were joined by tech giant Alibaba, which launched its Qwen smart glasses at the show — and having tried them, all I have to say is, Meta should watch its back. The Qwen glasses are among the first wearable devices Alibaba is building on top of its Qwen AI family of large language models, and the company brought two different models to the MWC.  The first pair, the Qwen S1 specs, have a heads-up waveguide display etched into the lenses, and serve as a rival to Meta’s Ray-Ban Display model (minus the gesture control). My first impression of these AR glasses was that they were light and comfortable to wear — I wouldn’t have known that they were smart glasses by their weight alone. At the end of each arm are swappable batteries, which snap off easily so you can keep …

Did Alibaba just kneecap its powerful Qwen AI team? Key figures depart in wake of latest open source release

Did Alibaba just kneecap its powerful Qwen AI team? Key figures depart in wake of latest open source release

Published by skeptic

Alibaba’s Qwen team of AI researchers have been among the most prolific and well-regarded by international machine learning community — shipping dozens of powerful generalized and specialized generative models starting last summer, most of them entirely open source and free. But now, just 24 hours after shipping the open source Qwen3.5 small model series—a release that drew public praise from Elon Musk for its “impressive intelligence density”—the project’s technical architect and several other Qwen team members have exited the company under unclear circumstances, raising questions and concerns from around the world about the future direction of the Qwen team and its focus on open source. The departure of Junyang “Justin” Lin, the technical lead who steered Qwen from a nascent lab project to a global powerhouse with over 600 million downloads, alongside two fellow colleagues — staff research scientist Binyuan Hui and intern Kaixin Li — marks a volatile inflection point for Alibaba Cloud and its role as an international open source AI leader. These three Qwen Team members announced their departures on X today, …

Alibaba’s Qwen tech lead steps down after major AI push

Alibaba’s Qwen tech lead steps down after major AI push

Published by skeptic

Alibaba’s Qwen AI project has lost one of its most visible technical leaders just a day after the Chinese tech giant unveiled its new Qwen 3.5 open-weight small models. Junyang Lin, a central technical leader on Alibaba’s Qwen team, said in a post on X on Tuesday that he was “stepping down” from the project, without elaborating. He joined Alibaba in July 2019 and became part of the Qwen team in April 2023, according to his LinkedIn profile. The abrupt departure, which drew strong reactions from colleagues and industry partners, comes as global competition among AI developers intensifies and companies race to build models rivaling those from OpenAI, Google, and Anthropic. Alibaba’s Qwen family of models has emerged as one of China’s most prominent open-weight AI efforts, with recent releases posting benchmark results that often rival systems from leading U.S. developers. The Chinese tech giant introduced the model in April 2023 and opened it to public use that September after receiving regulatory clearance. Alibaba introduced its Qwen 3.5 Small Model series on Monday, with four …

Alibaba’s Qwen 3.5 397B-A17 beats its larger trillion-parameter model — at a fraction of the cost

Alibaba’s Qwen 3.5 397B-A17 beats its larger trillion-parameter model — at a fraction of the cost

Published by skeptic

Alibaba dropped Qwen3.5 earlier this week, timed to coincide with the Lunar New Year, and the headline numbers alone are enough to make enterprise AI buyers stop and pay attention. The new flagship open-weight model — Qwen3.5-397B-A17B — packs 397 billion total parameters but activates only 17 billion per token. It is claiming benchmark wins against Alibaba’s own previous flagship, Qwen3-Max, a model the company itself has acknowledged exceeded one trillion parameters.  The release marks a meaningful moment in enterprise AI procurement. For IT leaders evaluating AI infrastructure for 2026, Qwen 3.5 presents a different kind of argument: that the model you can actually run, own, and control can now trade blows with the models you have to rent. A New Architecture Built for Speed at Scale The engineering story underneath Qwen3.5 starts with its ancestry. The model is a direct successor to last September’s experimental Qwen3-Next, an ultra-sparse MoE model that was previewed but widely regarded as half-trained. Qwen3.5 takes that architectural direction and scales it aggressively, jumping from 128 experts in the previous …

So Long, GPT-5. Hello, Qwen

So Long, GPT-5. Hello, Qwen

Published by skeptic

On a drizzly and windswept afternoon this summer, I visited the headquarters of Rokid, a startup developing smart glasses in Hangzhou, China. As I chatted with engineers, their words were swiftly translated from Mandarin to English, and then transcribed onto a tiny translucent screen just above my right eye using one of the company’s new prototype devices. Rokid’s high-tech spectacles use Qwen, an open-weight large language model developed by the Chinese ecommerce giant Alibaba. Qwen—full name 通义千问 or Tōngyì Qiānwèn in Chinese—is not the best AI model around. OpenAI’s GPT-5, Google’s Gemini 3, and Anthropic’s Claude often score higher on benchmarks designed to gauge different dimensions of machine cleverness. Nor is Qwen the first truly cutting-edge open-weight model, that being Meta’s Llama, which was released by the social media giant in 2023. Yet Qwen, and other Chinese models—from DeepSeek, Moonshot AI, Z.ai, and MiniMax—are increasingly popular because they are both very good and very easy to tinker with. According to HuggingFace, a company that provides access to AI models and code, downloads of open Chinese …