🧠 DeepSeek is Back

Today’s Highlights:

• 🧠 DeepSeek V3.2 Rivals GPT-5 and Gemini 3 Pro

• 🎤 Gradium Raises $70M to Build Lightning-Fast AI Voices

• 🚗 Nvidia Unveils First VLM for Self-Driving AI

• 👑 Microsoft Muscles Into the AI Enterprise Crown

• 📈 Runway Gen 4.5 Takes #1 Spot in AI Video 

• 🤝 OpenAI + Accenture Bring AI Agents to Corporates 

• 💥 Kling Omni Launch Week Sparks “Omniverse” Hype

^{Read Time: 4 mintue}

You can (easily) launch a newsletter too

This newsletter you couldn’t wait to open? It runs on beehiiv — the absolute best platform for email newsletters.

Our editor makes your content look like Picasso in the inbox. Your website? Beautiful and ready to capture subscribers on day one.

And when it’s time to monetize, you don’t need to duct-tape a dozen tools together. Paid subscriptions, referrals, and a (super easy-to-use) global ad network — it’s all built in.

beehiiv isn’t just the best choice. It’s the only choice that makes sense.

Start free today. No credit card required

LATEST AI NEWS

DEEPSEEK

🧠 DeepSeek V3.2 Rivals GPT-5 and Gemini 3 Pro

Chinese AI startup DeepSeek has returned to the spotlight with two new models—V3.2 and V3.2-Speciale—that directly challenge OpenAI’s GPT-5 and Google’s Gemini 3 Pro on logic, math, and code benchmarks.

The Details:

• V3.2 is now the default model on DeepSeek’s site and app, optimized for reasoning-first tasks and agentic AI use.

• V3.2-Speciale claims GPT-5 level performance, matching Gemini 3 Pro on complex problem-solving benchmarks.

• It earned gold-tier results in the 2025 International Math and Informatics Olympiads, overtaking Google and OpenAI’s previous leads.

• On tests like AIME and CodeForces, V3.2-Speciale topped every other model, including GPT-5 High and Gemini 3 Pro.

• While it beat GPT-5 High on Humanity’s Last Exam, it still ranked just behind Gemini 3 Pro.

Why it matters:

After a quiet phase, DeepSeek is back in the AI race with serious firepower. V3.2-Speciale’s elite benchmark scores show China’s top labs are catching up—and in some areas, pulling ahead.

GRADIUM

🎤 Gradium Raises $70M to Build Lightning-Fast AI Voices

Paris-based AI voice startup Gradium has emerged from stealth with a massive $70M seed round to take on industry giants like OpenAI and ElevenLabs with ultra-low-latency, multilingual AI voice models.

The Details:

• Founded in Sept 2025 by ex-Google DeepMind researcher Neil Zeghidour, spun out of French AI lab Kyutai.

• Raised $70M from FirstMark, Eurazeo, DST Global, Eric Schmidt, and Xavier Niel.

• Models focus on ultra-fast audio responses for real-time use cases like AI agents and voice apps.

• Launches with support for 5 languages: English, French, Spanish, German, and Portuguese.

• Faces stiff competition from major labs (OpenAI, Meta, Mistral) and startups like ElevenLabs, but aims to win on latency and expressiveness.

Why it matters:

As AI shifts from text to voice, real-time responsiveness becomes critical. Gradium’s massive backing and multilingual launch show Europe’s serious play in the AI voice race. Expect it to become a key player in voice-first agents and entertainment.

NVIDIA

🚗 Nvidia Unveils First VLM for Self-Driving AI

At NeurIPS 2025, Nvidia launched Alpamayo-R1, the first reasoning-capable vision language model (VLM) for autonomous driving — a major step toward AI-powered vehicles that understand and act like humans on the road.

The Details:

• Alpamayo-R1 is based on Nvidia’s Cosmos-Reason model, enabling vehicles to “see + think” before acting.

• It’s designed to support Level 4 autonomy, where cars drive themselves in defined areas without intervention.

• Nvidia also released the Cosmos Cookbook on GitHub: step-by-step tools, data prep, and training guides for physical AI developers.

• The VLM enables nuanced driving decisions using both image and text input, bringing common sense to navigation.

• All resources are available on GitHub and Hugging Face to support open-source collaboration in robotics and AVs.

Why it matters:

Physical AI is Nvidia’s next frontier. With Alpamayo-R1, they aim to power the “brains” behind future robots and self-driving cars. This could set the standard for how machines perceive and reason in the real world.

AI TUTORIAL

📈 Create Scroll-Stopping Product Ads with Make.ad

Make.ad uses AI to transform your product photos into polished, attention-grabbing ad visuals—no design skills or photoshoots needed.

How to Use:

• Visit make.ad and sign up for a free account

• Upload a product image or paste your product URL

• Enter a tagline or prompt describing the look you want

• Let AI instantly generate branded, ad-ready mockups and scenes

• Preview, tweak visuals to match your brand, then download your ads

👉 Try it free today at make.ad — elevate your product visuals in seconds!

MORE AI & TECH NEWS

• Microsoft is embedding Copilot across Office and launching tools like WorkIQ, Agent 365, and Foundry to lead in enterprise AI. It’s positioning itself as the core of corporate AI infrastructure.

• Apple’s AI head John Giannandrea is retiring; Amar Subramanya (ex-Microsoft/Google) will take over as Apple struggles with poor Apple Intelligence reception while competitors invest heavily in AI compute

• Kling AI kicked off Omni Launch Week, teasing its “Omniverse”—a bold step into high-end video creation. Creators are hyped, but some note persistent issues like lip-sync glitches.

• Runway’s Gen 4.5 model now tops the Video Arena leaderboard, praised for realistic motion, physics, and camera work—beating Google Veo 3 and OpenAI Sora 2 Pro.

• Alibaba, ByteDance, and DeepSeek are training AI models in Southeast Asia to bypass U.S. chip export rules. DeepSeek is also working with Huawei on alternative chip solutions.

• OpenAI & Accenture are partnering to scale AI agent adoption in big enterprises via ChatGPT Enterprise and automation programs—training thousands of staff.

• AI-generated music is dominating charts and streaming platforms. Most listeners can’t distinguish it from human-made songs, sparking concerns around ethics, job loss, and the fading value of human stories in music.

NEW AI TOOLS TO 10X YOUR PRODUCTIVITY

📊 Rose AI - Cloud platform for financial data analysis & visualization

🎬 DupDub - AI tool for fast multimedia content creation

🗣 Gliglish - AI app for interactive language learning

📚 ClassPoint AI - Generate quiz questions from slides instantly

🎤 MyVocal AI - Rapid voice cloning & text-to-speech platform

⚡ Fal.ai - Build apps with generative media AI models

✂️ Cutout.Pro - All-in-one AI suite for photo & video editing

AI MEME OF THE DAY

SPONSOR US

AI Insights newsletter is read by thousands of AI and Tech professionals/enthusiasts around the world.

Get in touch to get your product seen today!

Or email us at: [email protected]

THANK YOU FOR READING

FEEDBACK

We'd love to hear your feedback or any interesting thoughts you have!

Please share by replying to this email.