Large Language Models

Alibaba has launched its Qwen3.5 large language model series, featuring enhanced reasoning and native multimodal capabilities. The open-weight version offers flexibility for developers, while a hosted version is available on Alibaba Cloud. With 397 billion parameters and support for 201 languages, Qwen3.5 aims to compete with global AI leaders and addresses the growing trend of AI agents capable of autonomous task execution, amidst intense domestic competition from companies like ByteDance and Zhipu AI.

Chinese AI stocks surged on Thursday as companies launched advanced AI models and policymakers emphasized accelerating adoption. Zhipu AI and MiniMax saw significant gains after releasing new large language models and AI agent tools. This innovation push, aimed at narrowing the gap with U.S. competitors, also boosted infrastructure providers like UCloud Tech. The enthusiasm for pure-play AI firms contrasts with mixed performance from larger tech giants, but government support and a capital-efficient approach suggest China’s growing presence in the AI sector.

Knowledge Atlas Technology (Zhipu), a Chinese AI startup, debuted on the Hong Kong Stock Exchange after a $558 million IPO. The company, recognized as an “AI tiger,” is a leading developer of large language models and aims to compete globally. Despite U.S. restrictions impacting its AI model training capabilities, Zhipu plans to allocate a significant portion of its IPO funds to R&D. Another AI firm, MiniMax, is also expected to go public soon.

Anthropic is reportedly in late-stage talks for a $10 billion funding round, valuing the company at $350 billion. Coatue and GIC are leading the investment, signaling strong investor confidence. Founded by former OpenAI executives, Anthropic focuses on safe and ethical AI development with its Claude models. This funding would solidify its position against competitors like OpenAI and Google in the escalating AI arms race, with previous backing from Amazon, Microsoft, and Nvidia.

Kilo Code, an AI‑coding startup founded by former GitLab CEO Sid Sijbrandij and Scott Breitenother, raised an $8 million seed round. Its platform integrates with IDEs like VS Code and Cursor, offering a multi‑model API that processed over 3 trillion tokens in a month. Early adopters such as Plug & Pay report that 80% of their developers now rely on Kilo Code, cutting multi‑day tasks to a single day. With a $1,000 right of first refusal from GitLab and growing demand for “vibe coding,” the company is poised for further VC interest and possible acquisition.

AWS announced an AI‑enabled “DevOps Agent” that helps enterprises pinpoint and resolve system outages faster by ingesting data from tools like Datadog and Dynatrace. In preview, the service assigns multiple AI agents to test hypotheses, delivering root‑cause reports and remediation steps before engineers join. A pilot with Commonwealth Bank cut investigation time to under 15 minutes. The launch reflects cloud providers’ shift toward AI‑driven operations tools, a market projected to exceed $3 billion by 2028 as firms seek to lower costly downtime.

.AI spending drove two‑thirds of US GDP growth in H1 2025, prompting warnings of market froth. Yet corporate AI investment hit $252.3 bn in 2024, shifting focus from “whether” to “how” to spend. Only 5 % of firms profit from AI; they allocate >20 % of digital budgets, scale early, pursue transformative redesigns, and embed strong governance. Building proprietary LLMs is prohibitive, so diversifying across hyperscalers and alternative architectures mitigates supply risk. Success hinges on clear ROI use cases, organizational readiness, and proactive risk management, turning AI into a sustainable business‑transformation engine despite valuation volatility.

Google plans to double its AI server capacity every six months, potentially increasing it 1000-fold in 4-5 years. This expansion, backed by strong financials and a $93 billion capital expenditure forecast, reflects Google’s confidence in AI’s long-term value. Google emphasizes that infrastructure investment drives revenue, citing its cloud operations. Advances in TPUs and LLMs enhance efficiency. Industry experts agree that robust IT infrastructure is crucial for successful AI deployment, as inadequate systems hinder AI performance. Major technology providers are investing heavily in AI infrastructure to deliver scalable AI solutions.

A new review of 445 LLM benchmarks raises concerns about their validity and the reliance of enterprises on potentially misleading data for AI investment decisions. The study highlights weaknesses in benchmark design, including vague definitions, lack of statistical rigor, data contamination, and unrepresentative datasets. It urges businesses to prioritize internal, domain-specific evaluations over public benchmarks, focusing on custom metrics, thorough error analysis, and clear definitions relevant to their unique needs to mitigate financial and reputational risks.

Wikipedia founder Jimmy Wales is skeptical of Elon Musk’s Grokipedia, citing concerns about the reliability of Large Language Models (LLMs) used to generate its content and potential bias. Wales argues that LLMs are prone to errors and fabricating sources, unlike Wikipedia’s community-driven accuracy. He defends Wikipedia’s reliance on mainstream sources against Musk’s claims of “woke bias.” While not dismissing AI’s potential entirely, Wales believes current LLMs are inadequate for building trustworthy knowledge repositories and worries about the rise of AI-generated misinformation.