Morning Brief

• Tencent launches QClaw
• XPeng P7 introduces Max version and three new color options
• Xiaomi releases MiMo-V2-Pro model
• Midjourney unveils V8 model
• Firefox to introduce mascot “Kit”
• FIFA partners with YouTube for World Cup content
• Horizon Worlds to discontinue VR version

---

Tencent launches QClaw

On March 18, Tencent announced the public beta launch of QClaw, alongside a new version update that upgrades the WeChat entry into a mini program, enabling users to upload or receive files from their computers. A new “Inspiration Square” has also been introduced, featuring preset common tasks and skills, allowing users to execute actions with one click without writing commands.

QClaw is a local AI agent one-click launch package developed by Tencent PC Manager team, based on the OpenClaw framework. Its core highlight is the integration of both WeChat and QQ entry points, allowing users to remotely control their local computers by simply sending natural language commands within chat windows—without needing to open a dedicated client. It supports tasks such as file processing, data analysis, and browser automation.

The platform supports both Windows and macOS, integrates mainstream Chinese large models including Kimi, MiniMax, GLM, and DeepSeek, and connects to the ClawHub skill ecosystem, offering more than 5,000 predefined skills. Source

---

XPeng P7 introduces Max version and three new color options

On March 18, XPeng Motors held a product launch event, announcing three new exterior color options for the P7 series—Stellar Red, Midnight Black, and Nebula White—bringing the total number of available colors to ten. At the same time, the P7 Max version was introduced, offering three configurations: 702 Max, 820 Max, and 750 AWD Max, with prices ranging from RMB 203,800 to RMB 243,800.

The new model offers an intelligent upgrade package, allowing users to choose different configurations of Turing AI chips as well as second-generation VLA and VLM models. Two new interior color options—Gravity Red and Lava Gray—have also been added, along with “Stellar Knight” and “Dark Knight” styling packages. Source

---

Xiaomi releases MiMo-V2-Pro model

On March 19, Xiaomi unveiled its flagship foundation model for agent scenarios, Xiaomi MiMo-V2-Pro. According to the company, the model has over 1 trillion total parameters, with 42 billion activated parameters, and adopts a hybrid attention architecture. It supports up to 1 million tokens of context and is designed for high-intensity agent workflows, long-horizon planning, and complex tool usage in real-world scenarios.

Xiaomi stated that MiMo-V2-Pro ranks eighth globally and second in China on the Artificial Analysis leaderboard. It has demonstrated end-to-end task execution capabilities in agent frameworks such as OpenClaw and Claude Code, capable of completing complex workflows and precise tool calls without human intervention. The company claims its overall performance surpasses Claude Sonnet 4.6 and approaches Opus 4.6, while its API pricing is only about one-fifth of the latter.

In terms of capabilities, MiMo-V2-Pro continues the Hybrid Attention mechanism of its predecessor, increasing the ratio from 5:1 to 7:1, scaling up parameters while maintaining inference efficiency. Through expanded compute during post-training across broader agent tasks, the model extends beyond dialogue and code generation into full task execution. Xiaomi also revealed that the anonymous model previously released on OpenRouter under the name Hunter Alpha was an early internal test version of MiMo-V2-Pro.

MiMo-V2-Pro is now available via API, priced at $1 per million input tokens and $3 per million output tokens for contexts up to 256K, and $2 input / $6 output for contexts up to 1M tokens. The model has also been integrated into MiMo Studio, WPS Lingxi, Xiaomi Browser, and the mobile agent product miclaw. Source

---

Midjourney unveils V8 model

On March 18, the Midjourney team opened early access to the V8 model in its user community, inviting users to test the model and provide feedback. According to Midjourney, V8 shows significant improvements in understanding complex prompts, delivers better image detail and overall consistency, and further enhances text rendering capabilities. Generation speed has also increased by approximately five times compared to previous versions.

Alongside the V8 release, Midjourney has upgraded its web interface, introducing a more natural conversational creation workflow, a new Grid Mode for easier viewing of image sets, and a redesigned sidebar for parameter controls to reduce visual clutter. In terms of features, V8 supports multiple aspect ratios at launch, as well as parameters such as –chaos, –weird, –exp, and –raw, while remaining compatible with V7 personalization settings, mood boards, and style references.

The new version also introduces a native 2K output mode (–hd) and a –q 4 mode to improve visual consistency. Midjourney noted that Relax mode is not yet supported and is currently awaiting new server clusters. At present, tasks using –hd, –q 4, style references, and mood boards render at approximately one-quarter the speed of standard tasks and are billed at four times the cost.

Midjourney emphasized that V8 is a brand-new model with its own strengths and limitations, and may require different prompting techniques. Users are encouraged to use longer and more specific prompts, combined with personalization, style references, and mood boards to fully leverage V8’s capabilities, while also using the web-based rating system to help improve the model. Source

---

Firefox to introduce mascot “Kit”

On March 17, Mozilla announced that Firefox will retire its iconic “fox wrapped around the globe” logo, which has been associated with the browser for over 20 years, and introduce a new mascot named Kit.

In its blog post, Mozilla stated that the internet has entered a new era, and users need a new guide that conveys warmth and familiarity. Kit is designed to fill this role as a unified global identity. Conceptually, Kit is defined as a pure representation of Firefox—it is neither the traditional fox nor a Himalayan red panda. The character was created through a collaboration between Mozilla, creative agency JKR, and illustrator Marco Palmieri.

Kit will appear in Firefox’s onboarding experience, feature announcement pop-ups, and settings confirmation screens, and will also be used across Firefox’s website, blog, and social media platforms.

Mozilla also clarified that Kit was entirely handcrafted by human designers through hundreds of detailed refinements—such as tail movement, textures, and gradients—and is not generated by AI. Users are advised not to mistake it for a chatbot or AI assistant. Source

---

FIFA partners with YouTube for World Cup content

On March 18, YouTube announced a partnership with the Fédération Internationale de Football Association (FIFA), under which YouTube will serve as an official content partner platform for the 2026 World Cup. During the tournament, YouTube will provide match highlights, archival footage, and a dedicated creator program for global content creators.

According to current arrangements, rights-holding broadcasters will be allowed to livestream the first 10 minutes of each match on their YouTube channels, and may also publish selected full matches, extended highlights, behind-the-scenes footage, short-form videos, and on-demand content. FIFA will also open access to portions of its historical World Cup archives, including classic matches and iconic moments.

The collaboration will also include a global YouTube creator program, granting selected creators access to exclusive interviews and behind-the-scenes content during the tournament. The 2026 World Cup is scheduled to take place from June 11 to July 19 across the United States, Canada, and Mexico, featuring an expanded format of 48 teams and a total of 104 matches. Source

---

Horizon Worlds to discontinue VR version

On March 18, Meta confirmed that its metaverse platform Horizon Worlds will gradually phase out support for VR devices, with the VR version set to be discontinued and removed from the Quest Store on June 15.

According to Meta’s current timeline, starting March 31, the Quest Store will no longer display user-created Horizon Worlds environments and events, and users will be unable to access multiple in-world locations—including Horizon Central, Events Arena, Kaiju, and Bobber Bay—via VR headsets. By June 15, the Horizon Worlds app will be completely removed from Quest devices, after which users will only be able to access the platform in 2D mode via the Meta Horizon app on iOS or Android.

At the same time, the recently introduced Hyperscape Capture testing feature will also be discontinued. Meta stated that while users will still be able to capture and view Hyperscape content in the future, social features such as sharing, invitations, and multiplayer experiences will no longer be supported. Source