Tencent QClaw Receives a Major Upgrade: DeepSeek-V4-Pro Integrated for Immediate Use by AI Experts

Tencent Cloud officially announced today that its Agent platform QClaw has released version v0.2.14. This update is defined as the most significant version upgrade to date for the project, not only achieving diversification in the underlying model but also greatly lowering the barrier for users to use AI through a new framework and functional modules.

Comprehensive Evolution of Core Architecture and Models

In this update, QClaw first integrated the Hermes framework, allowing users to create and run Hermes-type Agents. This change enables users to schedule two different Agent kernels within a single application, significantly enhancing flexibility in handling complex tasks.

Regarding the underlying models, QClaw introduced a flexible switching mode. In addition to integrating Tencent's self-developed "fast-slow thinking" hybrid expert model, Huyuan Hy3 preview, it also supports the latest open-source models DeepSeek-V4-Pro, KIMI-K2.6, and GLM-5.1. Users can seamlessly switch between top-tier large models according to their needs.

New Experience in Expert Plaza and Remote Collaboration

To enhance usability, the original "Inspiration Plaza" has been upgraded to "Expert Plaza," which includes more than 100 industry AI experts. Users do not need to learn complex prompt techniques; they just need to select the corresponding expert and describe their needs to quickly obtain documents, code, or analysis reports.

Sub-millimeter Precision Alignment: Xiaomi Open Sources the Full Post-Training Process of VLA Large Model

Xiaomi has recently open-sourced the real-world post-training process of its vision-language-action large model, Xiaomi-Robotics-0, promoting the development of embodied intelligence. The team enabled the robot to master precise earphone storage and other complex tasks using only about 20 hours of task data, demonstrating the ability to quickly learn complex skills.

Anthropic Launches Project Deal: Claude Completes 186 Autonomous Transactions, Totaling Over $4,000

Anthropic quietly launched its internal experiment 'Project Deal' on April 24, 2026, showcasing the autonomous negotiation and trading capabilities of AI agents in real markets. The experiment took place in the Slack office market, where the Claude model represented 69 employees to execute buy and sell operations, facilitating 186 transactions among over 500 items, with a total transaction value exceeding $4,000. Technically, parallel market tests were used to compare the performance of different model specifications.

Upload a picture and instantly generate a 3D world. Ling Guang App brings the world model to mobile devices first

Ant Ling Guang App launches the 'World Model' feature, becoming the industry's first AGI product on mobile devices. Users can upload a picture and explore a 3D world on their phone for up to 60 seconds, supporting game-style perspective control, with exploration starting in seconds. This is the first time the world model has been implemented on the edge, featuring minute-level long-term consistency and real-time interaction capabilities.

Hot Trend: Bay Area Mansion Supports Exchange of AI Giant's Shares, Seller Claims AI Configuration Is Too Low

Investment banker Storm Duncan listed a 13-acre mansion in Mill Valley, San Francisco, and publicly requested that buyers pay the purchase price with shares of the AI unicorn Anthropic instead of cash. This move aims to balance asset allocation and highlight the value of tech equity in high-end transactions. Duncan even created a LinkedIn page for the property to attract attention from the technology and investment communities, showcasing the emerging trend of equity-for-property exchanges.

DeepSeek API Input Cache Price Dramatically Reduced: 1/10 of the Initial Price

DeepSeek announced a significant reduction in the price of its entire series of API input cache to 1/10 of the initial price. The input cache price for V4-Pro is now reduced to 0.1 yuan per million Tokens, and after a limited-time promotion, it is only 0.025 yuan, far lower than the overseas competitor GPT-5.5Pro. This move aims to attract more developers and enterprises by offering extreme cost-effectiveness, marking a new stage in domestic AI cost control.