ChatGPT's agent mode didn't make the cut. That’s how tough it was to crack our Top 25 list of AI tools and releases this year.

NotebookLM on the other hand? It won tool of the year. Again. Back to back.

(Dropping Nano Banana powered slides/video overviews and unlimited Studio generations turned NotebookLM into …. whatever comes after powerhouse.)

We JUST dropped our annual ranking of the 25 most impactful AI releases of 2025 on today's Everyday AI. And y'all, some of these results are gonna make you question everything you thought you knew about your AI strategy.

The tools busy executives SHOULD be using? 

Probably not the ones making headlines.

Your competitors are already using tools and modes you barely touch. And if you're not running multiple deep research sessions daily, you're already behind.

Time to fix that.

Let’s do (and abbreviated) breakdown. 👇

In Alphabetical order, here’s our Top 25 AI tools and releases of 2025:

1. Canva Visual Suite 2.0 🎨

This thing tries to turn Canva into a full multimedia creation suite with natural language prompts and the Magic Studio features. The image models and coding stuff aren't gonna blow your mind if you've got access to better tools. But that magic text grabbing feature is kinda clutch for quick fixes when you're on deadline.

2. ChatGPT Atlas 🌐

OpenAI's agentic browser that syncs with your ChatGPT memories both ways. So the browser knows what you've been working on and vice versa.

You can literally have an agent navigate the web for you, clicking through interfaces and doing QA on your workflows. The UI/UX has some quirks, but for scoping use cases alongside Perplexity's Comet, it's solid.

Related Episode: Ep 637: ChatGPT’s New Agentic browser: Hands on with OpenAI’s Atlas

3. ChatGPT Deep Research 🔬

An autonomous research agent that goes through dozens of sources and delivers a comprehensive cited report in five to 20 minutes.

If you're not running at least a couple of these daily for personalized learning, you are 100% gonna be behind your competition. Fire and forget research, but make it actually personalized to your specific business context.

Related Episode: Ep 470: OpenAI’s Deep Research Updates – What’s new and how to make it work for you

4. Claude Code 💻

A terminal-based coding agent that understands entire code bases and executes development tasks autonomously. It can read full code context, write, debug, and manage git repos.

But here's the thing. People are realizing you can use it for non-coding tasks too. It's gone viral in the last few weeks, which is weird because it's been out for months.

5. Claude Opus 4.5 🧠

Anthropic's most intelligent model and the first to exceed 80% on the SWE Bench verified benchmark. It's got great computer use capabilities, meaning it can operate graphical interfaces like a human.

Desktop automation. Web browsing. Terminal. Virtual computers. This is probably why we're getting closer to agents doing our jobs while we just orchestrate them.

Related Episode: Ep 662: Opus 4.5: New king of the AI hill or just a niche model for coders?

6. Cursor 2.0 ⚡

This AI-native code editor update brings composer and multi-agents to the forefront with tasks completing four times faster than before.

You can run up to eight agents in parallel, pick the best result, and plan with one agent while building with another. Great for large code base refactoring and teams wanting concurrent agent workflows.

7. DeepSeek v3/v3.2/R1 Family 🇨🇳

Chinese open source models that match a lot of frontier US models at dramatically lower costs with more transparent chain of thought reasoning.

Real talk here. There's a reason certain states and countries have banned it, so probably shouldn't be using it if you're a legitimate US business. Still made the list because it pushed everyone else to ship faster.

8. Genspark 🤖

A multi-agent workspace that distributes tasks across specialized agents working in parallel. Research, coding, analysis, and creative agents all at once.

You're essentially getting a team of AI specialists instead of one assistant, which means multiple perspectives on complex problems. They just had a big partner announcement with Microsoft at their conference.

Related Episode: EP 613: AI Agents: From automation to super agents. 10 AI Agents you should know in 2025

9. Gemini 3 Flash ⚡

A ridiculously good model that's technically free if you're using Google's AI mode instead of those crappy AI overviews.

On third party benchmarks, it's actually a top five model in the world. And it's the "small" version of Gemini 3 Pro. If you're searching anything on Google, click into AI mode every single time. It's one extra click. Just do it.

10. Gemini 3 Pro 👑

The number one model in the world on both LM Arena and Artificial Analysis. Multimodal by default. Video, audio, images with equal depth reasoning.

There's a million token context window on the API side and it actually understands the physics in video, not just transcripts. Google had a rough 2024 but they head-down shipped and crushed everyone in 2025.

Related Episode: Ep 656: Inside Gemini 3: What’s new and what it unlocks for your business with Google’s Logan Kilpatrick

11. Gemini Canvas Mode 🎨

An interactive workspace for creating and editing documents, code, and apps in real time. Feels like it's been around forever even though it dropped in 2025.

Use it for live document editing where you highlight certain things for co-editing instead of having the AI change random paragraphs. But the real unlock is rendering code. HTML, React prototypes, infographics, quizzes. Right there.

Related Episode: Ep 554: No-Code Interactive AI: How to use Canvas, Artifacts for business growth

12. GPT Image 1.5 🖼️

OpenAI's latest dedicated image model with improved quality and speed.

On LM Arena blind taste tests, it actually beats Nano Banana Pro for a lot of use cases. Way better than the original GPT image that went viral for cutesy stuff but wasn't actually that good. Best for marketers, advertisers, and communication professionals who need reliable image gen.

13. GPT 5.2 Pro 🏆

The model you go to when you need something right. Literally just yesterday it solved a physics problem humans couldn't crack.

Don't use the instant or auto versions because that's the 24th best model in the world. Always use GPT 5.2 Thinking or Pro. It's so good that deep research usage might actually go down because you can tap into canvas mode and other features while using this model.

14. Lovable 💜

A platform that builds full stack web applications from natural language with autonomous multi-step building. Front end, back end, database, deploys, everything.

They added Lovable Cloud with auth baked in so you don't have to duct tape backend services anymore. Best for nontechnical founders who have that app idea or companies paying $50k annually for software they barely use.

15. Manus 🤚

The general purpose super agent that can deliver complete work products from goal descriptions with up to 100 agents researching simultaneously on their own virtual computers.

They reportedly hit $100 million annual recurring revenue in eight months. Astronomical. Meta just acquired them for over $2 billion. Great for data analysis, resume screening, travel planning. Anything you'd give an assistant.

Related Episode: EP 613: AI Agents: From automation to super agents. 10 AI Agents you should know in 2025

16. ChatGPT Pulse 📱

A proactive daily briefing where ChatGPT researches overnight and delivers a personalized morning update by going through your email and calendar.

It's currently only for $200 a month users and honestly the feedback loop wasn't great during testing. But it signals the next step of AI. Proactive assistance without you asking. For casual users, probably solid. For power users, scheduled agents work better.

17. Microsoft 365 Copilot Agent Modes 💼

An AI assistant embedded throughout Microsoft 365 with autonomous capabilities including Memory Work IQ that maintains persistent context and learns performance.

The agent mode completes multi-step tasks across Word, Excel, PowerPoint, Outlook, and Teams autonomously with Copilot Studio access. Best for Microsoft-native enterprises who want to delegate complex workflows.

Related Episode: Ep 621: Microsoft’s new Agent mode: Microsoft marketing or Future of Work in Windows?

18. Nano Banana Pro 🍌

The first image model where literally nobody can tell if it's AI or not.

Not photographers. Not AI experts. Nobody.

Best in class text rendering that's finally solved, plus it's a reasoning model so it understands what's happening and what should happen in the photo. You can upload 14 different references and combine things for marketing assets, infographics, product mockups, and localized visuals.

19. NotebookLM Studio Updates 🏆

Our tool of the year for the second year in a row.

Personalized reports, customized templates, unlimited deep dive podcasts, different versions of overviews, plus video and slides powered by Nano Banana Pro. It's grounded only to what you give it so companies worried about LLMs can stop worrying. The fact that PowerPoint product managers have to compete with this thing? Brutal.

Related Episode: Episode 652 - NotebookLM Updates

20. Perplexity Comet ☄️

An agentic browser from Perplexity that syncs Chrome profiles, bookmarks, passwords, and extensions just like Atlas.

For tasks that don't need graphical interface clicking, it's faster than Atlas and does better at long range tasks in testing. Great for research across many sites, price comparisons, data collection. Anything requiring multiple browser tabs.

Related Episode: Episode 614 - Perplexity Comet: Five Business Use Cases

21. Replit Agent 3 🚀

A browser-based AI agent that builds, deploys, and hosts applications entirely in the cloud with natural language straight to deployed app.

No local setup required. Shaq literally made a full app using just his voice in their recent ad. Best for learning to code, deploying simple apps fast, rapid prototyping, or developing without a local environment.

22. Runway Gen 4.5 🎬

Professional grade AI video generation with Act One that captures facial and body motion from your webcam and transfers it to AI characters.

Can't get the right facial expression on your ad? Literally act it out and the avatar matches you. The multi motion brush for selective animation and quality jump from Gen 4 make it solid for marketers, filmmakers, and commercial production.

23. Sora 2 🎥

OpenAI's video model with native synchronized audio including dialogue, synced lips, sound effects, and music in clips up to 25 seconds.

You gotta sign up via mobile app which is technically a social network slash brain rot, but then you can use it on desktop. Character cameos let you upload yourself and do anything. Scary possibilities aside, it's great for short form content and social video.

Related Episode: Ep 623: Sora 2: AI TikTok Brain Rot or Your Company’s Secret Creative Weapon?

24. Suno v5 🎵

The AI music generator with a massive quality jump in natural vocals and better instrument separation where you can control each layer with text prompts.

There was actually an artist breaking onto billboard charts before people found out it was just someone using Suno. That's how good it is. Probably better than 99% of musicians unless you're touring or on the charts. Sorry not sorry.

Related Episode: Episode 207 - Suno CEO Interview

25. Veo 3.1 📹

Google's flagship video model that was first to generate synchronized audio natively including ambient noise, music, dialogue, and sound effects in up to 4K resolution.

The physics understanding is through the roof, and beginning frame to end frame generations are bonkers. Best for video content creation, marketing, and anyone wanting complete videos without needing a huge editing team.

Google won 2025.

NotebookLM is the tool of the year. Again.

And if you're using AI like it's 2024, your competitors are already lapping you.