Doubao Pro Goes Live: Taps into Doubao 2.1 Pro Model with Three-Tier Pricing
On June 24, Doubao officially rolled out its professional version, powered by the new 2.1 series of large models. Pro users get full access to the office task mode driven by the Doubao 2.1 Pro model, while free users can still test-drive the Doubao 2.1 Turbo model within specific usage limits.
Here’s how the pricing works: It’s broken down into three clear tiers. The Standard Plan is 68 yuan per month (billed continuously), offering over five times the quota of the free version for features like Expert Mode and Office Tasks. If you need more power, the Enhanced Plan jumps to 200 yuan/month, giving you four times the quota of the standard tier. For the heavy lifters, the Premium Plan costs 500 yuan/month, unlocking a massive 10x quota boost compared to the standard option.
Masayoshi Son Says Arm Will Enter Chip Manufacturing, Eyes 10x Growth Potential
At the annual shareholders meeting on June 24, SoftBank Group Chairman Masayoshi Son dropped some major news: his UK-based chip design firm, Arm, is evolving from just a designer into a full-fledged chip provider, and he’ll be personally overseeing the manufacturing side. Son predicts that the coming AI era will revolve around CPUs, emphasizing that Arm still has “more than 10 times room to grow.” He also touched on SoftBank’s investment in Intel, noting that while it was criticized back then, the company is now generating trillions of yen in profit based on market value.
Qualcomm Confirms Acquisition of AI Startup Modular
On June 24, Qualcomm issued a statement confirming they’ve reached an agreement to acquire AI startup Modular. The deal is expected to wrap up in the second half of 2026.
According to reports, as part of this transaction, Qualcomm plans to issue up to 19.2 million shares of its common stock to Modular’s shareholders. Based on Qualcomm’s closing price last time around, this deal values the acquisition at approximately $3.92 billion.
Groq Raises $650 Million to Scale AI Inference Cloud, Targets 200MW Capacity by Late 2027
AI chip startup Groq announced on June 22 (local time) that it has successfully closed a new growth round of $650 million, led by Disruptive and Infinitum, with participation from existing investors. The fresh capital will fuel upgrades to their existing data center infrastructure, including the deployment of NVIDIA’s latest LPX systems. Their ambitious goal? To expand total installed capacity to 200 megawatts by the end of 2027. On the leadership front, Groq welcomed several new executives: Alan Rice, formerly from xAI (now SpaceX AI) and Meta’s data centers, joins as COO. Meanwhile, Sinclair Schuller and Rakesh Malhotra are set to take over as CTO and Chief Product Officer, respectively, starting in July.
Masayoshi Son Reveals SoftBank Mass Producing Robots, Aims to Become World No. 1 via ABB Deal
Speaking at the shareholder meeting on June 24, SoftBank Chairman Masayoshi Son hinted at breakthroughs in physical AI applications. “We’ve already started mass-producing robots in a factory,” he revealed, adding that an official launch is imminent and everyone will be surprised. Son outlined a vision where SoftBank aggregates top-tier robot companies across various verticals to become the “overwhelmingly number one robot company in the world.” He specifically mentioned plans to complete the acquisition of the robotics division from Swiss industrial giant ABB by 2026.
Alibaba Cloud Introduces ‘Peak-Valley Tokens’ for QoderWork: Save Up to 80% on Qwen3.7 Usage
On June 24, Alibaba Cloud announced a smart new feature for QoderWork called “Peak-Valley Tokens.” If you run tasks between 10 PM and 8 AM, you automatically get significant discounts, with the Qwen3.7-Max model dropping to just 20% of the regular price. This night-time discount applies across QoderWork, QoderDesktop, CLI, and other products. Users can simply schedule tasks during the day or submit long-running commands before bed, letting the Agent handle the execution overnight. By morning, your work is done, and you’ll have consumed only 20%-40% of the credits compared to daytime rates.
Jimeng AI Launches Native 4K Version for Seedance 2.0 VIP
As of June 23, the web version of Jimeng AI has officially enabled native 4K capabilities for its Seedance 2.0 VIP tier, supporting ultra-high-definition output. Unlike post-processing upscaling, native 4K preserves crucial details right from the source. Whether it’s individual hair strands, fabric textures, architectural lines, or material grain, high-density information is rendered more completely and clearly. This significantly boosts image detail, color depth, and lighting effects, making it perfect for high-end creative scenarios like film post-production, brand visual design, and advertising marketing.
