option

Discover quality AI tools

Bring together the world’s leading artificial intelligence tools to help improve work efficiency

Author ArthurYoung

Articles published by ArthurYoung

A total of 11 articles
September 20, 2026

DeepSeek clarified its API pricing, confirming off-peak rates for adjusted workdays and national holidays. This creates a ten-day period from September 20 to October 10 with discounted calls, including Mid-Autumn and National Day breaks. Additionally, its new V4.1 Flash model offers enhanced cost-efficiency, outperforming the flagship V4Pro while maintaining low off-peak prices of 0.02 yuan per million input tokens and 4 yuan per million output tokens.

DeepSeek clarified its API pricing, confirming off-peak rates for adjusted workdays and national holidays. This creates a ten-day period from September 20 to October 10 with discounted calls, including Mid-Autumn and National Day breaks. Additionally, its new V4.1 Flash model offers enhanced cost-efficiency, outperforming the flagship V4Pro while maintaining low off-peak prices of 0.02 yuan per million input tokens and 4 yuan per million output tokens.

DeepSeek clarified its API pricing, confirming off-peak rates for adjusted workdays and national holidays. This creates a ten-day period from September 20 to October 10 with discounted calls, including Mid-Autumn and National Day breaks. Additionally, its new V4.1 Flash model offers enhanced cost-efficiency, outperforming the flagship V4Pro while maintaining low off-peak prices of 0.02 yuan per million input tokens and 4 yuan per million output tokens.
August 21, 2026

Unitree Robotics founder Wang Xingxing said at the 2026 World Robot Conference that the company is advancing self-evolving physical AI robot models by using large AI models to auto-develop control code, which will be verified through evaluation and manual review. The system will incorporate simulation, real-world and human behavior data for continuous improvement as robot deployment scales. The firm also demonstrated robots evolving from single to group intelligence through multi-machine coordination, showing cluster tasks performed by humanoid, quadrupedal and wheel-legged models along with the new GD01 manned exoskeleton. Wang Xingxing noted the biggest hurdle for embodied AI is insufficient generalization ability (around 80% task completion via voice commands in unfamiliar settings) which could be achieved in 2 to 3 years or take 5 to 10 years otherwise, adding that current models struggle with correcting tactile errors in fine operations due to input output discrepancies.

Unitree Robotics founder Wang Xingxing said at the 2026 World Robot Conference that the company is advancing self-evolving physical AI robot models by using large AI models to auto-develop control code, which will be verified through evaluation and manual review. The system will incorporate simulation, real-world and human behavior data for continuous improvement as robot deployment scales. The firm also demonstrated robots evolving from single to group intelligence through multi-machine coordination, showing cluster tasks performed by humanoid, quadrupedal and wheel-legged models along with the new GD01 manned exoskeleton. Wang Xingxing noted the biggest hurdle for embodied AI is insufficient generalization ability (around 80% task completion via voice commands in unfamiliar settings) which could be achieved in 2 to 3 years or take 5 to 10 years otherwise, adding that current models struggle with correcting tactile errors in fine operations due to input output discrepancies.

Unitree Robotics founder Wang Xingxing said at the 2026 World Robot Conference that the company is advancing self-evolving physical AI robot models by using large AI models to auto-develop control code, which will be verified through evaluation and manual review. The system will incorporate simulation, real-world and human behavior data for continuous improvement as robot deployment scales. The firm also demonstrated robots evolving from single to group intelligence through multi-machine coordination, showing cluster tasks performed by humanoid, quadrupedal and wheel-legged models along with the new GD01 manned exoskeleton. Wang Xingxing noted the biggest hurdle for embodied AI is insufficient generalization ability (around 80% task completion via voice commands in unfamiliar settings) which could be achieved in 2 to 3 years or take 5 to 10 years otherwise, adding that current models struggle with correcting tactile errors in fine operations due to input output discrepancies.
August 20, 2026

Payment giant Stripe has finalized the acquisition of AI model routing platform OpenRouter for around $7.5 billion, a value more than five times higher than OpenRouter’s prior valuation of $1.3 billion following its Series B funding in May. OpenRouter provides developers with a unified interface to route requests to hundreds of AI models based on user needs, handling over 10 trillion tokens daily and serving 10 million global users with integrated models from 400 to 500 mainstream providers. The merged entity will operate independently under OpenRouter’s existing name and mission, keeping its product roadmap unchanged while prioritizing user interests. Industry analysts say Stripe aims to integrate these routing capabilities into its payment system to manage token consumption in the growing agent economy and gain control over core elements of AI-driven economic operations, with the deal expected to close within weeks.

Payment giant Stripe has finalized the acquisition of AI model routing platform OpenRouter for around $7.5 billion, a value more than five times higher than OpenRouter’s prior valuation of $1.3 billion following its Series B funding in May. OpenRouter provides developers with a unified interface to route requests to hundreds of AI models based on user needs, handling over 10 trillion tokens daily and serving 10 million global users with integrated models from 400 to 500 mainstream providers. The merged entity will operate independently under OpenRouter’s existing name and mission, keeping its product roadmap unchanged while prioritizing user interests. Industry analysts say Stripe aims to integrate these routing capabilities into its payment system to manage token consumption in the growing agent economy and gain control over core elements of AI-driven economic operations, with the deal expected to close within weeks.

Payment giant Stripe has finalized the acquisition of AI model routing platform OpenRouter for around $7.5 billion, a value more than five times higher than OpenRouter’s prior valuation of $1.3 billion following its Series B funding in May. OpenRouter provides developers with a unified interface to route requests to hundreds of AI models based on user needs, handling over 10 trillion tokens daily and serving 10 million global users with integrated models from 400 to 500 mainstream providers. The merged entity will operate independently under OpenRouter’s existing name and mission, keeping its product roadmap unchanged while prioritizing user interests. Industry analysts say Stripe aims to integrate these routing capabilities into its payment system to manage token consumption in the growing agent economy and gain control over core elements of AI-driven economic operations, with the deal expected to close within weeks.
July 16, 2026

Megvii Intelligent partners with Samsung to integrate its MiniCPM series edge-side models into flagship smartphones. The on-device AI enables faster responses, privacy protection, and offline operation. Multiple flagship models launching soon, expected to enhance voice assistants, image recognition, and personalized recommendations.

Megvii Intelligent partners with Samsung to integrate its MiniCPM series edge-side models into flagship smartphones. The on-device AI enables faster responses, privacy protection, and offline operation. Multiple flagship models launching soon, expected to enhance voice assistants, image recognition, and personalized recommendations.

Megvii Intelligent partners with Samsung to integrate its MiniCPM series edge-side models into flagship smartphones. The on-device AI enables faster responses, privacy protection, and offline operation. Multiple flagship models launching soon, expected to enhance voice assistants, image recognition, and personalized recommendations.
July 7, 2026

Meta ran Project Cannes using outsourced staff posing as minors to stress-test competitor AI chatbots like ChatGPT and Gemini with over 45,000 high-risk prompts on suicide and self-harm. Meta calls it standard safety benchmarking, denying use of the data to train its own models, sparking debate on ethical AI testing boundaries.

Meta ran Project Cannes using outsourced staff posing as minors to stress-test competitor AI chatbots like ChatGPT and Gemini with over 45,000 high-risk prompts on suicide and self-harm. Meta calls it standard safety benchmarking, denying use of the data to train its own models, sparking debate on ethical AI testing boundaries.

Meta ran Project Cannes using outsourced staff posing as minors to stress-test competitor AI chatbots like ChatGPT and Gemini with over 45,000 high-risk prompts on suicide and self-harm. Meta calls it standard safety benchmarking, denying use of the data to train its own models, sparking debate on ethical AI testing boundaries.
July 2, 2026

AReaL 2.0, an open-source reinforcement learning infrastructure, launches to enable agents to continuously learn in real-world business scenarios. It records agent interactions, transforms them into training data, and supports safe online RL. Developed by Ant Group, Tsinghua, and HKUST, the project is now open-sourced on GitHub and linked to the PyTorch Foundation.

AReaL 2.0, an open-source reinforcement learning infrastructure, launches to enable agents to continuously learn in real-world business scenarios. It records agent interactions, transforms them into training data, and supports safe online RL. Developed by Ant Group, Tsinghua, and HKUST, the project is now open-sourced on GitHub and linked to the PyTorch Foundation.

AReaL 2.0, an open-source reinforcement learning infrastructure, launches to enable agents to continuously learn in real-world business scenarios. It records agent interactions, transforms them into training data, and supports safe online RL. Developed by Ant Group, Tsinghua, and HKUST, the project is now open-sourced on GitHub and linked to the PyTorch Foundation.
June 16, 2026

NVIDIA plans to issue at least $20 billion in bonds in seven tranches with maturities from 2 to 30 years, marking its first debt sale since 2021. The longest bonds yield 0.9 percentage points above U.S. Treasuries. JPMorgan, Morgan Stanley, and Goldman Sachs are underwriting the sale. Proceeds will fund general corporate purposes including refinancing. The move aligns with tech giants like Alphabet and Amazon using debt to expand AI computing capacity, underscoring the capital efficiency race in AI infrastructure.

NVIDIA plans to issue at least $20 billion in bonds in seven tranches with maturities from 2 to 30 years, marking its first debt sale since 2021. The longest bonds yield 0.9 percentage points above U.S. Treasuries. JPMorgan, Morgan Stanley, and Goldman Sachs are underwriting the sale. Proceeds will fund general corporate purposes including refinancing. The move aligns with tech giants like Alphabet and Amazon using debt to expand AI computing capacity, underscoring the capital efficiency race in AI infrastructure.

NVIDIA plans to issue at least $20 billion in bonds in seven tranches with maturities from 2 to 30 years, marking its first debt sale since 2021. The longest bonds yield 0.9 percentage points above U.S. Treasuries. JPMorgan, Morgan Stanley, and Goldman Sachs are underwriting the sale. Proceeds will fund general corporate purposes including refinancing. The move aligns with tech giants like Alphabet and Amazon using debt to expand AI computing capacity, underscoring the capital efficiency race in AI infrastructure.
June 2, 2026

French tech brand WIKO launched AI pet hardware Xingzai on June 1, priced at 499 yuan, sales start June 30. It features multimodal interaction (touch, shake, voice), sound localization, and deep integration with Huawei Xiaoyi large model for emotional companionship and mood diary. Supports Wi-Fi and 4G, 1900mAh battery, available in three colors. Pre-orders open.

French tech brand WIKO launched AI pet hardware Xingzai on June 1, priced at 499 yuan, sales start June 30. It features multimodal interaction (touch, shake, voice), sound localization, and deep integration with Huawei Xiaoyi large model for emotional companionship and mood diary. Supports Wi-Fi and 4G, 1900mAh battery, available in three colors. Pre-orders open.

French tech brand WIKO launched AI pet hardware Xingzai on June 1, priced at 499 yuan, sales start June 30. It features multimodal interaction (touch, shake, voice), sound localization, and deep integration with Huawei Xiaoyi large model for emotional companionship and mood diary. Supports Wi-Fi and 4G, 1900mAh battery, available in three colors. Pre-orders open.
May 12, 2026

OpenAI amended its bylaws to require an absolute majority of non-employee directors to remove CEO Sam Altman, significantly strengthening his position after the 2023 board challenge. This change makes his dismissal legally difficult and solidifies his control over the AI firm.

OpenAI amended its bylaws to require an absolute majority of non-employee directors to remove CEO Sam Altman, significantly strengthening his position after the 2023 board challenge. This change makes his dismissal legally difficult and solidifies his control over the AI firm.

OpenAI amended its bylaws to require an absolute majority of non-employee directors to remove CEO Sam Altman, significantly strengthening his position after the 2023 board challenge. This change makes his dismissal legally difficult and solidifies his control over the AI firm.
April 9, 2026

Dell CEO warns AI accelerator memory demand will surge 625x by 2028 versus 2023, driven by a 25-fold increase in per-unit capacity (to 2TB) and a 25-fold expansion in deployment scale. This creates unprecedented strain on the semiconductor supply chain. With new memory chip fab construction taking ~4 years and cautious industry expansion, shortages are expected to persist until 2028.

Dell CEO warns AI accelerator memory demand will surge 625x by 2028 versus 2023, driven by a 25-fold increase in per-unit capacity (to 2TB) and a 25-fold expansion in deployment scale. This creates unprecedented strain on the semiconductor supply chain. With new memory chip fab construction taking ~4 years and cautious industry expansion, shortages are expected to persist until 2028.

Dell CEO warns AI accelerator memory demand will surge 625x by 2028 versus 2023, driven by a 25-fold increase in per-unit capacity (to 2TB) and a 25-fold expansion in deployment scale. This creates unprecedented strain on the semiconductor supply chain. With new memory chip fab construction taking ~4 years and cautious industry expansion, shortages are expected to persist until 2028.
OR