Zo | zo.me

This week brought powerful new models, AI video breakthroughs, and custom AI agent frameworks.

From Google’s open-source Gemma 3 to OpenAI’s new AI agent-building tools, here’s everything you need to know. (Bookmark this for later!)

OpenAI unveiled a new framework for building custom AI agents.

Developers can now create task-specific AI agents with real-time web search and app integrations, unlocking smarter automation.

We're launching new tools to help developers build reliable and powerful AI agents. 🤖🔧

Timestamps:
01:54 Web search
02:41 File search
03:22 Computer use
04:07 Responses API
10:17 Agents SDK https://t.co/vY514tdmDz

OpenAI Developers

Zo is the OS for Human to Agent coordination

Zo is a group chat app and protocol powered by AI

Perplexity AI launched its official Windows app.

With voice dictation, keyboard shortcuts, and the latest AI models, Perplexity now brings its powerful AI search experience to the desktop.

Perplexity AI launched its official Windows app.

With voice dictation, keyboard shortcuts, and the latest AI models, Perplexity now brings its powerful AI search experience to the desktop.https://x.com/perplexity_ai/status/1899498357154107499

Google just dropped Gemma 3, a family of open-source AI models built on Gemini 2.0 tech.

With sizes ranging from 1B to 27B parameters, 128K-token context, and multimodal support, it's one of the most flexible open models yet.

Google just dropped Gemma 3, a family of open-source AI models built on Gemini 2.0 tech.

With sizes ranging from 1B to 27B parameters, 128K-token context, and multimodal support, it's one of the most flexible open models yet.https://x.com/googleaidevs/status/1899725682545967555

Manus AI teams up with Alibaba’s Qwen team to supercharge its AI agents.

This partnership optimizes Manus AI for China’s computing platforms, making it more powerful and efficient.

Manus AI teams up with Alibaba’s Qwen team to supercharge its AI agents.

This partnership optimizes Manus AI for China’s computing platforms, making it more powerful and efficient.https://x.com/China__Focus/status/1899681004295458913

Luma AI introduced Ray2 Flash, an upgraded text-to-video model that’s 3x faster and more affordable than its predecessor.

AI video generation is getting cheaper and better!

Luma AI introduced Ray2 Flash, an upgraded text-to-video model that’s 3x faster and more affordable than its predecessor.

AI video generation is getting cheaper and better!https://x.com/LumaLabsAI/status/1898056614684381218

Reka AI launched Flash 3 Reasoning, a 21B open-source model focused on efficiency and complex problem-solving.

It’s designed to balance performance and cost-effectiveness—a new contender in the AI reasoning space.

Reka AI launched Flash 3 Reasoning, a 21B open-source model focused on efficiency and complex problem-solving.

It’s designed to balance performance and cost-effectiveness—a new contender in the AI reasoning space.https://x.com/RekaAILabs/status/1899481289495031825

Tencent released Hunyuan-TurboS, offering enhanced speed and quality over previous iterations.

Designed for both creative and practical applications, it’s a solid step forward in multimodal AI.

Tencent released Hunyuan-TurboS, offering enhanced speed and quality over previous iterations.

Designed for both creative and practical applications, it’s a solid step forward in multimodal AI.https://x.com/TXhunyuan/status/1899105803073958010

Hedra launched Character 3, an omnimodal model for creating animated talking characters from text or audio.

It blends voice synthesis, lip-syncing, and expressive visuals, targeting video creators who want hyper-realistic animations.

Hedra launched Character 3, an omnimodal model for creating animated talking characters from text or audio.

It blends voice synthesis, lip-syncing, and expressive visuals, targeting video creators who want hyper-realistic animations.https://x.com/hedra_labs/status/1897699010632466469

Google added native image generation and editing to Gemini 2.0 Flash.

Now available via Google AI Studio, it allows for text-prompted image creation and multi-turn edits—no separate model like Imagen needed.

Google added native image generation and editing to Gemini 2.0 Flash.

Now available via Google AI Studio, it allows for text-prompted image creation and multi-turn edits—no separate model like Imagen needed.https://x.com/GoogleDeepMind/status/1899896275652202927

Freepik partnered with Google to integrate Veo 2, unlocking image-to-video generation. Now, you can transform static images into high-quality, dynamic AI videos—a game-changer for creators!

Freepik partnered with Google to integrate Veo 2, unlocking image-to-video generation. Now, you can transform static images into high-quality, dynamic AI videos—a game-changer for creators!
https://x.com/freepik/status/1899741722156138830

That’s a wrap on this week’s AI agent developments!

Which release are you most excited about? Drop your thoughts below! 👇

(And hit Follow for weekly AI updates 🚀)

OpenAI unveiled a new framework for building custom AI agents.

Developers can now create task-specific AI agents with real-time web search and app integrations, unlocking smarter automation.https://x.com/OpenAIDevs/status/1899531225468969240