Z.ai has released its GLM-5.3 Flash AI model, a multimodal model that can handle text, image, and video inputs. This model is part of the GLM-5 series and has shown competitive performance on various benchmarks. It is a mixture-of-experts model with 320B total parameters and 18B active per token.
Alibaba's Qwen team has also released Qwen3.8-Flash-Next, a multimodal mixture-of-experts model with 176B parameters. This model features a hybrid Gated DeltaNet and Qwen Sparse Attention architecture to address long-context inference bottlenecks.
In other news, Cathie Wood's Ark Invest has purchased 93,290 shares of Cerebras Systems, valuing the transaction at approximately $17.2 million. Cerebras recently unveiled its CS-4 system, which claims to outperform traditional GPU-based systems in AI inference.
Meanwhile, investors may consider taking profits in AI stocks ahead of Nvidia earnings and uncertainty around future data-center spending. Morgan Stanley recommends trimming AI positions to rebalance portfolios.
In acquisitions, Brokurz has acquired MLS Ingest, a real estate technology company, and Nooks has acquired AI sales-coaching startup FullyRamped.
The global military power dynamics are changing with the integration of AI and robotics. China is rapidly advancing in industrial robotics, while the US retains a lead in autonomy software.
Amazon's AI training facility, VGT3, scans and destroys books for AI training data. The facility receives shipments of books, which are then scanned and destroyed.
Key Takeaways
- Z.ai releases GLM-5.3 Flash AI model, a multimodal model with 320B total parameters.
- Alibaba's Qwen team releases Qwen3.8-Flash-Next, a multimodal mixture-of-experts model with 176B parameters.
- Cathie Wood's Ark Invest purchases 93,290 shares of Cerebras Systems for $17.2 million.
- Morgan Stanley recommends trimming AI positions ahead of Nvidia earnings.
- Brokurz acquires MLS Ingest, a real estate technology company.
- Nooks acquires AI sales-coaching startup FullyRamped.
- China advances in industrial robotics, while US leads in autonomy software.
- Amazon's AI training facility scans and destroys books for AI training data.
- GLM-5.3 Flash is available for download on Hugging Face and released under an MIT license.
- Qwen3.8-Flash-Next achieves significant performance improvements on various benchmarks.
Z.ai Unveils GLM-5.3 Flash AI Model
Z.ai has released its new GLM-5.3 Flash AI model, which is a multimodal model that can handle text, image, and video inputs. The model is part of the GLM-5 series and has shown competitive performance on various benchmarks, including DeepSWE and AutomationBench. GLM-5.3 Flash is a mixture-of-experts model with 320B total parameters and 18B active per token. It has been served on a large cluster of Chinese AI chips, demonstrating improved performance and efficiency. The model is available for subscribers of GLM Coding Plan and its weights are publicly available.
Z.AI's Ox Alpha Model Revealed
The Ox Alpha model, previously unknown, has been revealed to be developed by Chinese lab Z.AI. The model, also known as GLM-5.3-Flash, is a multimodal model that can handle text, image, and video inputs. It has shown competitive performance on various benchmarks, including coding and agentic workloads. Z.AI plans to release the model's weights, making it open for anyone to use. The model's release could change the calculus for companies currently pricing frontier access as a premium product.
Z.ai Releases GLM-5.3-Flash Model
Z.ai has released GLM-5.3-Flash, a natively multimodal model in the GLM-5 series. The model has a 1,048,576-token context window and can handle image and video inputs. It is a mixture-of-experts model with 320B total parameters and 18B active per token. GLM-5.3-Flash is available for download on Hugging Face and is released under an MIT license.
Qwen3.8-Flash-Next Multimodal Model
Alibaba's Qwen team has released Qwen3.8-Flash-Next, a multimodal mixture-of-experts model with 176B parameters. The model features a hybrid Gated DeltaNet and Qwen Sparse Attention architecture to address long-context inference bottlenecks. Qwen3.8-Flash-Next achieves significant performance improvements on various benchmarks and can be run on NVIDIA GB300 NVL72 hardware.
Alibaba's Qwen3.8-Flash-Next Model
Alibaba's Qwen team has released Qwen3.8-Flash-Next, a multimodal mixture-of-experts model with 125B backbone and 51B N-gram embedding table. The model activates 6B parameters per token and is designed for cost per token. Qwen3.8-Flash-Next is an early preview of the Qwen4 architecture.
Military AI and Robotics Arms Race
The global military power dynamics are changing with the integration of AI and robotics. China is rapidly advancing in industrial robotics and has installed 295,000 industrial robots in 2024. The US retains a lead in autonomy software, but China's advancements in robotics and AI could pose a challenge. The UN has established a scientific panel on AI, and a resolution on lethal autonomous weapons received 156 votes in favor.
Brokurz Acquires MLS Ingest
Brokurz has acquired MLS Ingest, a real estate technology company that connects and standardizes multiple listing service property data. The acquisition advances Brokurz's strategy of bringing brokerage operations, property data, automation, and artificial intelligence into a single technology platform for real estate businesses.
Cathie Wood's Investment in Cerebras Systems
Cathie Wood's Ark Invest has purchased 93,290 shares of Cerebras Systems, valuing the transaction at approximately $17.2 million. Cerebras recently unveiled its CS-4 system, which claims to outperform traditional GPU-based systems in AI inference. The investment reflects Ark Invest's strategic positioning in the rapidly evolving tech landscape.
Amazon's AI Training Facility
An Amazon employee revealed that the company's AI training facility, VGT3, scans and destroys books for AI training data. The facility receives shipments of books, which are then scanned and destroyed. The employee described the process as disorganized, with books being thrown into shuttles.
Investors Consider Taking Profits in AI Stocks
Investors may consider taking profits in AI stocks ahead of Nvidia earnings and uncertainty around future data-center spending. Morgan Stanley recommends trimming AI positions to rebalance portfolios.
Nooks Acquires FullyRamped
Nooks has acquired AI sales-coaching startup FullyRamped to expand its agent platform. The deal aims to provide Nooks' customers with more comprehensive sales coaching and training tools.
Top 5 Japanese Electronics Stocks for AI Exposure
Morgan Stanley identified five Japanese electronics manufacturers as preferred picks for exposure to AI infrastructure spending. The top pick is Murata, followed by Ibiden, TDK, Hirose, and Niterra.
Dual-Dimensional LLM Framework for Automated Item Similarity Analysis
A new study proposes a dual-dimensional framework for Automated Item Similarity Analysis (AISA) powered by Large Language Models (LLMs). The framework aims to improve the detection of incidental content redundancy in large-scale assessments.
Sources
- Z.AI Reveals Ox Alpha Is GLM 5.3 Flash, Competes With Claude Opus 4.8 & GPT 5.6 Terra On Benchmarks
- Mystery Ox Alpha Model Revealed To Be From Chinese Lab Z.AI
- Z.ai Releases GLM-5.3-Flash: A 320B-A18B Natively Multimodal MoE With a 1M-Token Context
- Experiment with Qwen3.8-Flash-Next 176B Model on NVIDIA GB300 NVL72 for Agentic Coding
- Alibaba’s Qwen Team Releases Qwen3.8-Flash-Next: A 125B Multimodal MoE With 6B Active Parameters Previewing the Qwen4 Architecture
- The Military AI and Robotics Arms Race, Explained
- Brokurz Acquires MLS Ingest to Expand Real Estate AI and Data Infrastructure
- Cathie Wood Loads Up on Cerebras Systems Stock as AI Chip Race With Nvidia Heats Up — Ark Keeps on Dumping AMD Shares
- Inside the Warehouse Where Amazon Scans and Destroys Books for AI Training
- Why investors should consider taking profits in AI stocks now
- Dealroom.co | Nooks buys AI sales-coaching startup FullyRamped to expand agent platform
- Top 5 Japanese Electronics Stocks for AI exposure, According To Morgan Stanley
- A Dual-Dimensional LLM Framework for Automated Item Incidental Content Similarity Analysis in Large-Scale Assessments
Comments
Please log in to post a comment.