TechNode

A research team involving miHoYo’s Honkai AI researchers and university collaborators has open-sourced GameScaling, a family of models that play games by generating keyboard and mouse actions from visual observations. The release includes eight checkpoints ranging from 800 million to 27 billion parameters, built through continued pretraining of Qwen3.5 vision-language models. The models use current…

DeepSeek has completed a $15 billion private funding round, nearly doubling its initial target following strong investor demand, The Wall Street Journal reported on October 22, citing people familiar with the matter. The Chinese AI developer is also accelerating preparations for a potential Shanghai listing early next year, according to the report. The financing would…

Editor’s note: This article was written by Lucia, a TechNode reporter. For businesses adopting AI, the question is no longer simply which model to use. It is also where AI should run, how much it will cost, and whether sensitive data can remain under the company’s control. PeakVision AI’s answer is to bring more AI…

If you have used AI video tools, you may have noticed an interesting trend: generating a visually impressive video is becoming easier, but turning it into a commercial-ready production is still far from simple. What happens when a subtitle is wrong? What if product data needs to be updated? If a brand logo appears in…

Editor’s note: This article was written by Lucia, a TechNode reporter. Building a more capable AI model is only half the battle. The other half is making it affordable to run at scale. DeepSeek announced a V4.1 Flash update aimed at faster task completion and shorter outputs, alongside an open-source framework for more efficient inference.…

DeepSeek released DeepInfer, the inference framework behind its large-scale model services, on October 20. The framework supports Nvidia GPUs and Huawei Ascend processors and provides capabilities including distributed inference, speculative decoding and structured output. DeepSeek said the release is intended to help developers and businesses improve throughput and latency while reducing computing costs and energy…

Chinese robot hardware startup YITA announced on October 20 that it has raised more than RMB 100 million across its seed and seed-plus rounds. Investors included Galbot, Fourier, Matrix Partners China, CCB International, SenseTime Guoxiang Capital and the L2F founder fund. Lighthouse Capital served as the company’s incubator and exclusive financial adviser. Founded in May…

Tencent has opened Yuanbao Search to external developers and businesses through Tencent Cloud, turning the search infrastructure used by its own AI products into a commercial service. Announced on October 20, the offering combines publicly available web information with Tencent’s content ecosystem and is designed to provide information retrieval for large language models and AI…

As more AI hardware focuses on voice assistants, information processing, and workplace productivity, one Chinese startup is taking a different approach by using AI to help people preserve their life experiences. Shanghai-based Vibe Technology has introduced Souvi, an outdoor AI memory device designed for activities such as hiking, camping, travel, and urban exploration. By combining…

ACE Robotics, founded by SenseTime co-founder Xiaogang Wang, released its ACE-Brain-0.6 embodied AI model on October 19. Built through supervised fine-tuning of Qwen3.5-9B, the model focuses on identifying objects, locating them in images and determining where a robot can interact with them. The company said it trained the model using approximately 1 million spatial-grounding samples.…

Tencent has recently introduced Lüpaopao, a free short-drama service inside WeChat, offering AI-generated series alongside live-action and animated titles. The mini-program adds another distribution channel to China’s expanding market for serialized mobile entertainment. ByteDance’s Hongguo reached 168 million daily active users in July, up 107% year on year, according to QuestMobile figures cited in Nomura…

DeepSeek released V4.1 Flash-1017 on October 17, updating its Flash model with improvements in reasoning, coding, agent tasks and visual understanding. The model weights are available on Hugging Face under the MIT license. The company said the update reduces task execution time and output length, allowing users to complete work faster and at lower cost.…

Rokid has partnered with Tencent Mobility Services to launch ride-hailing through its AI glasses. Announced on October 16, the service lets users state a destination by voice and receive vehicle details and estimated arrival times through the glasses. The integration connects Rokid’s assistant with Tencent’s travel-service agent using the Agent-to-Agent protocol. Tencent provides functions including…

Hangzhou-based robot AI startup Gewu Tongzhi has completed an angel funding round of nearly RMB 100 million, according to reports published on October 18. Investors include Hongfu Asset Management, Peak Investment, Hangzhou Capital and Hangzhou Talent Group. Founded in 2026, the company develops AI for dexterous robotic manipulation, combining world models with learning from physical…

DeepSeek has launched Search and Fetch APIs, opening its proprietary search engine to developers building AI agents and applications that retrieve external information. Search returns structured web results, while Fetch extracts a webpage’s title and content as Markdown. Both services cost RMB 5 per 1,000 successful requests. Failed requests, empty search results and timeouts are…

Editor’s note: This article is republished from Jumpstart, a TechNode Group publication. In 2023, AI software was DataCanvas’s business. By the first half of 2026, it was almost absent from its sales. The change is unusually stark. AI foundational software services generated 76.5% of the Chinese company’s revenue in 2023. Three years later, their contribution…

AI video generation is no longer just about producing a visually impressive clip. For creators looking to use AI to make short films, advertisements or even feature films, the harder questions are whether characters can remain consistent, whether shots can unfold as intended, and whether a video can tell a coherent story from beginning to…

Shenzhen-based robot AI developer Ruoyu Technology has completed a nine-figure RMB Pre-A funding round led by Sunway Communication, 36Kr reported on October 15. Other investors include Jinmu Capital, Qindao Capital, Huanghai Capital and Zhongchuang Capital. Founded in 2023 and incubated by Harbin Institute of Technology, Shenzhen, Ruoyu develops AI systems that help robots perceive environments,…

Ant Group has released the weights of Ling-3.1-flash, making its AI model available for developers to deploy on their own infrastructure. The model has approximately 560 billion total parameters, activates about 25 billion per token and supports a context window of up to one million tokens. The release follows the model’s earlier introduction and includes…

Xiaomi has open-sourced MiDashengLM-Spatial, an audio-language model that combines general sound understanding with spatial awareness. The company released model weights and inference code under the Apache 2.0 license. Built on Xiaomi’s MiDashengLM audio model, the system processes binaural audio to identify sound sources and reason about their positions and spatial relationships. It retains capabilities such…