According to a BlockBeats report, Tencent's Hunyuan team has taken a full week of collective leave after releasing the Hy4 preview. The team's usual workspace, previously occupied almost daily, has been largely empty this week, with the break tied to the latest model release. Hy4 preview was open-sourced on August 28, marking Tencent's largest open-source model to date. It carries 770 billion total parameters, 49 billion activated parameters, and supports a context length exceeding 1 million. In the months before the launch, the team pushed the project forward under sustained high-intensity effort. Once the version went live and won internal approval, the company arranged a structured rest period. In an industry where high-intensity R&D has become standard practice, many model teams tend to move directly into the next iteration after shipping a release. Tencent's decision to insert a centralized pause after reaching a milestone is therefore relatively rare in the industry.

B.AI, the AI computing platform, said on August 31 that its cumulative token throughput has officially surpassed 5.74 trillion (5T+), a historic milestone for the platform's ongoing free-access program. The announcement arrived as the campaign for frontier large language models keeps running at full tilt. Under massive concurrent requests and demanding Agent workloads, B.AI's industrial-grade infrastructure has demonstrated stable capacity across the board, the company said. Six flagship models are now available on the platform: GLM-5.3-Flash (Ox Alpha), Qwen3.8-Flash, DeepSeek-V4-Flash, DeepSeek-V4-Flash-Vision-Exp, Tencent Hy3, and Xiaomi MiMo-V2.5 — all with zero entry barriers, unlimited usage, and no cost. The lineup covers a broad set of tasks, including code generation, ultra-long text processing, multimodal visual understanding, and Agent workflow deployment. Developers who sign in from today onward can call the entire suite at zero cost, free of charge and with unlimited compute integrated directly. B.AI pitches the setup as genuine "computing freedom" for every developer and AI power user alike.

Alibaba launched Qwen3.7-Plus with input priced at $0.40 and output at $1.60 per million tokens, about 80% cheaper than Qwen3.7-Max. The tradeoff is clear: the model is offered only through Alibaba Cloud APIs and Qwen Chat, with no open weights.

Brent A. Yorgey of Hendrix College used an open letter to question the tech industry his students are entering, laying out six career and ethics principles centered on moral limits, deep work, and putting people ahead of profit.

BlockDAG unveiled an AI model, World Cup promo, miner shipments, MiCA update, and listing delays — but BDAG dropped 30% in a week, hitting $0.00003726.

Andrej Karpathy, an OpenAI founding member now on Anthropic’s pretraining team, said he often switches to voice input and talks to a large language model for about 10 minutes when a task is too complex to explain cleanly by typing. In a post on X on Tuesday, July 21, he described the approach as a “long ramble session,” where ideas can come out unordered and messy. His view is that LLMs are unusually good at reconstructing what a user is actually trying to do from a long, incoherent monologue, then returning a cleaner version of that intent. Karpathy said he sometimes opens by telling the model he is using speech recognition and apologizing in advance for transcription errors. In other cases, he turns the exchange into a short interview and lets the model ask follow-up questions to pull out scattered details. The point, he suggested, is not voice input itself but the amount of context delivered up front. Instead of spending time compressing a thought into a polished prompt, he hands the model the whole bundle and lets it reorganize it. The post also sparked debate on X. Supporters said the method accounts for much of their prompt usage, with some claiming 5x to 8x gains in output efficiency. Critics raised a different concern: if the model takes over the work of clarifying ideas, writing and organization skills could weaken over time.

CanIRun.ai uses browser-based hardware detection to estimate which local LLMs a machine can run and at what speed, but users are also questioning its coverage, accuracy, and privacy implications.

According to OpenRouter, China's AI model usage exceeded the US for the fifth week in a row, reaching 12.96 trillion tokens (+31.48%), while US models saw only 0.76% growth. Alibaba's Qwen3.6 series dominates the top six, highlighting a structural shift in global AI compute consumption.