Zo | zo.me

This week in AI has seen massive breakthroughs in multimodal models, ChatGPT 4o update goes full Ghibli style and Deepseek stunned everyone with their latest model topping the charts.

Here’s your weekly roundup in AI that you can’t miss! (Bookmark this for later👇)

Google’s Gemini 2.5 Pro is here.

It’s already the #1 model on the LMArena leaderboard, excelling in reasoning, math, coding, and science. With a 1M-token context window, it’s a powerhouse for developers.

Introducing Gemini 2.5, our most intelligent AI model. 

Our first release, an experimental version of 2.5 Pro, unlocks state-of-the-art performance in math and science. 🔥

Learn more 🧵 https://t.co/aoe7egliJb

Google

Zo is the OS for Human to Agent coordination

Zo is a group chat app and protocol powered by AI

Google’s Gemini 2.5 Pro is here.

It’s already the #1 model on the LMArena leaderboard, excelling in reasoning, math, coding, and science. With a 1M-token context window, it’s a powerhouse for developers.https://x.com/Google/status/1904579652816941506

Anthropic redesigned the Claude interface.

With a cleaner and more refined UI, building and interacting with Claude models is now even more seamless.

We've done some spring cleaning.

The Claude interface is now more refined, thanks to your feedback. https://t.co/2iY4C9uXpL

Anthropic

Anthropic redesigned the Claude interface.

With a cleaner and more refined UI, building and interacting with Claude models is now even more seamless.https://x.com/AnthropicAI/status/1905341566040113375

OpenAI introduced a new native image generator for GPT-4o.

It enables the creation of high-res images from text prompts and includes advanced editing tools.

Expect your feed to be flooded with creative visuals soon!

4o image generation has arrived.

It's beginning to roll out today in ChatGPT and Sora to all Plus, Pro, Team, and Free users. https://t.co/pFXDzKhh2t

OpenAI

OpenAI introduced a new native image generator for GPT-4o.

It enables the creation of high-res images from text prompts and includes advanced editing tools.

Expect your feed to be flooded with creative visuals soon!https://x.com/OpenAI/status/1904602845221187829

DeepSeek has upgraded its model to DeepSeek V3.

This version offers better reasoning capabilities than ever before.

Pro tip: Deselect the “DeepThink” button to maximize its potential.

🚀 DeepSeek-V3-0324 is out now!

🔹 Major boost in reasoning performance
🔹 Stronger front-end development skills
🔹 Smarter tool-use capabilities

✅ For non-complex reasoning tasks, we recommend using V3 — just turn off “DeepThink”
🔌 API usage remains unchanged
📜 Models are https://t.co/QVuPwCODne

DeepSeek

DeepSeek has upgraded its model to DeepSeek V3.

This version offers better reasoning capabilities than ever before.

Pro tip: Deselect the “DeepThink” button to maximize its potential.https://x.com/deepseek_ai/status/1904526863604883661

Alibaba unveiled LHM, the Large Animatable Human Reconstruction model.

It generates fully 3D animated human avatars from a single full-body photo and is open-source under Apache 2.0.

Alibaba just released LHM on Hugging Face

Large Animatable Human Reconstruction Model from a Single Image in Seconds https://t.co/XLMiijPCAb

Alibaba unveiled LHM, the Large Animatable Human Reconstruction model.

It generates fully 3D animated human avatars from a single full-body photo and is open-source under Apache 2.0.https://x.com/_akhaliq/status/1903208078155133134

Microsoft launched Researcher.

This AI tool combines:

- OpenAI’s deep research models
- Microsoft 365 Copilot’s orchestration capabilities

It’s designed for tackling complex, multi-step research projects.

summary_large_image

Our Researcher and Analyst agents are like having a highly skilled expert on call for you 24/7 across your work data and the web. Excited to bring reasoning to Microsoft 365 Copilot &amp; Copilot Studio today. https://t.co/1eC43EzQdi

Satya Nadella

Microsoft launched Researcher.

This AI tool combines:

- OpenAI’s deep research models
- Microsoft 365 Copilot’s orchestration capabilities

It’s designed for tackling complex, multi-step research projects.https://x.com/satyanadella/status/1904739853037887716

Reve launched Reve Image 1.0.

Fresh out of stealth, it’s now the top image generation model, outperforming giants like MidJourney and Google’s Imagen.

- Stunning photorealism
- Industry-best prompt accuracy
- Advanced text rendering

Halfmoon is Reve Image — and it’s the best image model in the world 🥇
(🔊) https://t.co/Zm1FzNQaFh

Reve

Reve launched Reve Image 1.0.

Fresh out of stealth, it’s now the top image generation model, outperforming giants like MidJourney and Google’s Imagen.

- Stunning photorealism
- Industry-best prompt accuracy
- Advanced text rendering https://x.com/reveimage/status/1904211082870456824

Ideogram 3.0 is live!

The update offers:
- Improved photorealism
- Enhanced text rendering
- Better language understanding

It also introduces style reference and random style features to unlock more creative possibilities.

Meet Ideogram 3.0 — stunning realism, creative designs, and consistent styles, all in one powerful model. And it's blazingly fast.

Now available to all Ideogram users for free. https://t.co/k8IzpKivN2

Ideogram

Ideogram 3.0 is live!

The update offers:
- Improved photorealism
- Enhanced text rendering
- Better language understanding

It also introduces style reference and random style features to unlock more creative possibilities.https://x.com/ideogram_ai/status/1904927717281456188

Qwen has dropped three new models:

- QVQ-Max: Built for high-quality image-text processing
- Qwen2.5-Omni-7B: A versatile model for general AI tasks
- Qwen2.5-VL-32B-Instruct: Specialized for multimodal image-text tasks

72B too big for VLM? 7B not strong enough! Then you should use our 32B model, Qwen2.5-VL-32B-Instruct!

Blog: https://t.co/2yx5MXsnCW
Qwen Chat: https://t.co/FmQ0B9tiE7
HF: https://t.co/A4A2VmOQ0w
ModelScope: https://t.co/k5fg0rToe2

This time, we further optimize this VLM with https://t.co/JwEHMCOv79

Qwen

Qwen has dropped three new models:

- QVQ-Max: Built for high-quality image-text processing
- Qwen2.5-Omni-7B: A versatile model for general AI tasks
- Qwen2.5-VL-32B-Instruct: Specialized for multimodal image-text tasks https://x.com/Alibaba_Qwen/status/1904227859616641534

Perplexity launched Answer Tabs.

This feature organizes search results into categories, making AI-driven search faster and more intuitive than ever.

Perplexity for ___________.

Discover more than ever with new answer tabs on Perplexity. Search for images, video, travel, shopping, and more—all in one place.

Available now on web. Coming soon to mobile. https://t.co/2TZSLp47ko

Perplexity

Perplexity launched Answer Tabs.

This feature organizes search results into categories, making AI-driven search faster and more intuitive than ever.https://x.com/perplexity_ai/status/1904566323201687848

A Real Estate Agent built with Firecrawl’s Deep Research API and Claude 3.7 Sonnet.

Developed by @ericciarla and team, this AI tool provides real-time apartment listings based on your location and preferences.

We built an AI Real Estate Agent with Firecrawl's Deep Research API! 🔥

Just enter your location and preferences to get real-time, up-to-date apartment listings instantly.

Powered by Claude 3.7 Sonnet and @firecrawl_dev. https://t.co/JuNQfT372I

Eric Ciarla (hiring)

A Real Estate Agent built with Firecrawl’s Deep Research API and Claude 3.7 Sonnet.

Developed by @ericciarla and team, this AI tool provides real-time apartment listings based on your location and preferences. https://x.com/ericciarla/status/1904203105727901860

MCPs and the Future of Agent-Native Apps

@zenorocha explores why email is the perfect proving ground for AI agents and how MCPs are reshaping APIs designed specifically for agents instead of humans.

An agent-first world is around the corner — where more actions will likely be taken by AI than by humans when building software.

@zenorocha, cofounder of @resend, and Yoko Li say this reality is shaping how developers build, and changing what user experience really means.

Great https://t.co/CGdNj2khQG

a16z

Build a team of MCP Agents to plan your next trip! 🌍✈️

Welcome to the future of seamless travel planning, where AI handles it all. Your AI-powered travel concierge awaits! 🙌 @Saboo_Shubham_

Build a team of MCP Agents to plan your next trip! 🌍✈️

Welcome to the future of seamless travel planning, where AI handles it all. Your AI-powered travel concierge awaits! 🙌 @Saboo_Shubham_
https://x.com/Saboo_Shubham_/status/1904938092508635511

That’s a wrap on this week’s AI breakthroughs!

Which of these releases caught your eye the most?

Follow us for more weekly updates on the most exciting developments in the world of AI agents!

MCPs and the Future of Agent-Native Apps

@zenorocha explores why email is the perfect proving ground for AI agents and how MCPs are reshaping APIs designed specifically for agents instead of humans.https://x.com/a16z/status/1903868805308137581