Stop Using Chatgpt For Everything: The Ai Models I Use For Research, Coding, And More (and Which I Avoid)

6 days ago

Follow ZDNET: Add america arsenic a preferred source on Google.

ZDNET's cardinal takeaways

Different AI models triumph astatine images, coding, and research.
App integrations often adhd costly AI subscription layers.
Obsessing complete exemplary type matters little than workflow.

The gait of alteration successful nan burgeoning generative AI world is blisteringly fast. It's often difficult to support up pinch everything, moreover if it's your full-time job. Readers show maine that 1 area they find peculiarly confusing is nan wide array of poorly-named AI models. What successful nan heck is nan quality betwixt GPT-5.1, Opus 4.5, Gemini 3, etc.? And why would you usage 1 complete nan other?

Also: Claude Code made an astonishing $1B successful 6 months - and my ain AI-coded iPhone app shows why

To beryllium honest, trying to afloat understand nan elaborate differences betwixt each of nan AI models will nonstop almost anyone screaming into nan woods. But it's reasonably easy, particularly pinch examples, to understand which models to take for different tasks. That's what we're going to do successful this article.

Keep successful mind that apps and AI models are different beasts. The exemplary is nan underlying AI motor that does nan intelligence processing. The exertion is nan instrumentality you use. Here's a diagram, created successful Google's Nano Banana Pro, that gets it mostly right.

But successful nan aforesaid measurement that different vehicles usage different kinds and brands of engines, different applications usage different models. For example, I gave nan nonstop aforesaid punctual that produced nan supra sketch to ChatGPT's image generator (on nan left) and Midjourney's image generator (on nan right), and we sewage these results:

As you tin see, ChatGPT's consequence is elemental and reasonably clear, whereas Midjourney went some overboard while besides simultaneously not really pursuing nan prompt. Midjourney is awesome for conceptual images (like my bunny extracurricular a edifice creation shown successful this article), but it's reasonably unspeakable astatine diagrams.

Also: The champion AI chatbots of 2025: I tested ChatGPT, Copilot, and others to find nan apical devices now

That said, ever support successful mind my friend Vinnie Moscaritolo's description: "AI isn't nan sharpest instrumentality successful nan box. Sometimes it gets it perfect, and sometimes it dives consecutive into nan rabbit spread of stupid for illustration it packed a luncheon for nan trip." Rabbit spread of stupid. Gotta retrieve that phrase.

The cost/profit facet

Most AI companies waste their models some arsenic APIs (which programmers tin incorporated into their applications) and arsenic chatbot applications. Keep successful mind that an AI chatbot isn't an AI connected its own. An AI chatbot for illustration ChatGPT is really a chat exertion (like a messaging app aliases Slack) that calls into an AI exemplary for intelligence work.

While location are immoderate really bully free AI chatbots, erstwhile you commencement utilizing them a lot, you're apt to bargain into 1 aliases much of nan astir $20/month plans each offers.

You've astir apt noticed really galore package vendors are now including AI capabilities. I've been a QuickBooks Online personification for galore years, but now QuickBooks is each astir pitching "upgrades" to waste you AI supplier subscriptions.

Some vendors let you to usage nan AI capabilities from nan paid chatbot plans you're already paying for wrong their apps. But astir exertion vendors want that further gross stream. Even if you already salary for AI models for illustration GPT-5.1 aliases Opus 4.5 done your chatbot plan, astir package vendors incorporating AI will complaint you their ain interest for entree to those models.

Also: 4 ways I prevention money connected my favourite AI instrumentality subscriptions - and you tin too

How you determine what to usage should beryllium based connected what you need. To thief retired pinch that decision, I'll spell complete nan models I usage for galore of nan different tasks I do for work. Keep successful mind that I seldom take nan exemplary first. I take nan task and application, and nan exemplary usually comes arsenic portion of that application.

Creating explainers

Models used: Gemini 3 and Gemini 2.5

A cleanable illustration of this is Google's NotebookLM, a instrumentality I added to my kit astir six months agone and now usage astatine slightest weekly.

NotebookLM is simply a notebook instrumentality integrated pinch a connection exemplary (hence, NotebookLM). But I don't usage it arsenic a notebook. While immoderate folks for illustration that capability, NotebookLM has a superpower: it creates audio explainers based connected nan root worldly it is provided with.

Also: I recovered an open-source NotebookLM replacement that's powerful, backstage - and free

You tin take whether you want consecutive audio aliases an audio way accompanied by slides. Mostly, I conscionable spell for nan consecutive audio. Whenever I get a peculiarly dense document, property release, aliases method paper, I provender it into NotebookLM. About 15 minutes later, nan app produces a chat exploring nan cardinal points and main issues of immoderate I've fed to it.

To beryllium clear, I ne'er usage that output straight successful my work. But it's a awesome measurement to get immoderate speedy triangulation connected nan large issues related to immoderate I'm moving on.

Google doesn't explicitly picture nan exemplary underlying NotebookLM, but it's surely a Gemini variant. Since Gemini 3 conscionable came out, astir of nan activity I've done pinch NotebookLM was apt generated pinch Gemini 2.5.

Identifying keywords

Models used: GPT-4o, GPT-5, GPT-5.1

Notice that I listed 3 models above. That's because OpenAI updated them. As they were updated, nan exemplary I utilized automatically upgraded arsenic well. That's why I don't want you to obsess connected exemplary versions. If you publication this successful 3 months, each nan models will beryllium different.

In this case, I'm utilizing a self-hosted web archiving exertion called Karakeep. Karakeep replaced Pocket arsenic my measurement of storing articles I've recovered for early usage successful research. One point that Karakeep does that Pocket ne'er could is create really awesome automatic keywords.

Karakeep does this utilizing nan API from OpenAI. Because nan self-hosted Karakeep is unfastened root and doesn't impact immoderate subscription fee, its doesn't return its AI taxation retired of fees. Instead, I usage an API cardinal provided by OpenAI, and salary OpenAI based connected my usage.

I person 24,594 items indexed successful Karakeep. Moving my listings from Pocket to Karakeep took astir 2 months, because Karakeep had to rescan each article. It besides called retired to OpenAI truthful nan AI could make keywords. That process costs maine astir $40.

Since nan original conversion, I person been billed $5 each mates of months. Well worthy it.

Coding

Models used: ChatGPT 5.1, GPT-5.1-Max, Claude Opus 4.5

I usage AI models successful 2 very different ways erstwhile coding. For projects wherever I'm dropping a chunk of codification into nan chatbot and asking what it does, what's incorrect pinch it, aliases why I'm getting an correction message, I haven't recovered thing amended than ChatGPT Plus utilizing GPT-5.1.

When I tested the free chatbots for coding prowess, Microsoft Copilot really hit ChatGPT. And erstwhile I later tested nan paid type of Claude successful a chatbot, Opus 4.5, it failed half my tests. But nan $20/month ChatGPT Plus, utilizing nan latest GPT, has reliably done a awesome job.

Also: The champion free AI for coding successful 2025 - only 3 make nan trim now

However, erstwhile I do agentic coding, I've recovered that OpenAI's Codex utilizing GPT-5.1-Max and Claude Code utilizing Opus 4.5 are astonishingly great. Agentic AI coding is erstwhile I hook up nan AIs to my improvement environment, fto nan AIs publication my full codebase, and past do substantial, multi-step tasks.

For example, I utilized Codex to constitute four WordPress plugin products for maine successful 4 days. Just recently, I've been using Claude Code pinch Opus 4.5 to build an full analyzable and blase iPhone app, which it helped maine do successful small sprints complete nan people of astir half a month. I spent $200 for nan month's usage of Codex and $100 for nan month's usage of Claude Code.

It does astonish maine that Opus 4.5 did truthful poorly successful nan chatbot experience, but was a superstar successful nan agentic coding experience, but that's portion of why we're looking astatine different models. AI vendors are still moving retired nan kinks from this nascent technology.

Notion databases

Models used: Claude, ChatGPT, Gemini

Despite my annoyance pinch Notion's prioritization of AI upsells complete basal merchandise functionality, I yet collapsed down and paid Notion's AI taxation for nan year. It doesn't telephone it that, of course. Instead, it offers very constricted AI proceedings capabilities pinch its $12/month Plus plan. You person to upgrade to nan $24/mo business scheme to get nan AI features for real. I saved a fewer bucks by buying nan twelvemonth astatine once, which made my effective monthly interest $20/month.

Also: Dreading AI occupation cuts? 5 ways to future-proof your profession - earlier it's excessively late

There are 2 things I usage Notion AI for. The first is searching and summarizing my article drafts. ZDNET doesn't let extracurricular AIs to entree its articles, but I often request to excavation backmost into my erstwhile articles to reappraisal what I've written.

I tin constituent Notion AI to my draught library, and it will scan nan full thing, not conscionable searching for articles, but summarizing what I've antecedently said. This is awesome for assignments for illustration "Summarize what I've written astir location labs by looking successful nan blogs database."

Note: I don't usage this AI-generated contented successful my articles. It's solely for readying purposes.

As you tin spot from nan screenshot above, I tin popular unfastened nan Notion AI model and inquire questions of it, correct while I'm writing. The article you spot successful nan screenshot is this article, still being drafted.

The different point Notion AI does for maine is to move elephantine lists into categorized databases. I did a task wherever I gathered together a ample database of AI-related programming tools, but I wanted to beryllium capable to prime and position devices by vendors and categories. I gave Notion nan list, and it created a database for me. It wasn't perfect, but it is usable.

Notion doesn't usage conscionable 1 model. It tends to jump betwixt Claude, ChatGPT, and Gemini (and their various models) depending connected what it thinks will do nan occupation champion and which exemplary tin do it for nan slightest cost.

Speech nickname

Model: Parakeet

I usage nan Mac's dictation characteristic rather often. My small pup likes to slumber connected my near shoulder. That's adorable, but it makes it somewhat difficult to type pinch 2 hands. So I usage my rodent pinch my correct hand, and often dictate mini blocks of matter utilizing nan Mac's autochthonal dictation feature.

Recently, I've begun exploring dictation programs that connection much precocious reside nickname and study capabilities. One specified instrumentality is Paraspeech. Paraspeech offers 2 features I find rather compelling.

Also: I've been testing AI contented detectors for years - these are your champion options successful 2025

First, you salary conscionable once. There's nary subscription fee. This is imaginable owed to its 2nd feature: nan AI runs connected your ain machine. When you instal Paraspeech, it downloads a variety of nan Nvidia Parakeet reside nickname model. The use of this is that your reside doesn't get sent to nan cloud. It conscionable stays connected your machine.

This illustration shows really an AI exemplary tin disagree considerably from different models. This exemplary runs correct connected my ain machine. It's designed specifically to do reside recognition. And it's included wrong nan application, alternatively than called by it. That allows nan developer to complaint a one-time interest without having to salary usage fees backmost to nan AI vendor. It whitethorn besides springiness nan developer a income advantage, because paying erstwhile is thing rather charismatic to astir consumers. We each salary subscription fees erstwhile we person to, but cipher likes them.

Deep investigation

Models: GPT-5.1 Thinking (in ChatGPT Pro)

Deep investigation is nan process by which nan AI spends a batch of clip figuring retired answers and providing information. It often uses a ton of AI resources, truthful it costs much to access. Back erstwhile I had nan $200/mo Pro subscription for my programming, I besides utilized Deep Research to do immoderate reasonably astonishing things.

My favourite illustration was erstwhile I asked Deep Research to create a trading briefing archive from root code. I utilized a alternatively agelong prompt, truthful I'm conscionable going to summarize. What I told it was that there's a reasonably modular process wherever engineering managers little trading and merchandise managers connected what a portion of exertion does.

Also: Stop accidentally sharing AI videos - 6 ways to show existent from clone earlier it's excessively late

Marketing managers past usage that briefing, which includes features, tech tidbits (think Apple presentation) and usage cases, arsenic a instauration to thief them nutrient each their trading materials and activities.

But I didn't person an engineering manager. Instead, I told Deep Research to publication each 12,000+ lines of codification from nan root codification files, and from them, intuit what nan products did, and from that, constitute nan briefing documents. The results were astonishing. The AI sewage a fewer insignificant things wrong, but they were easy to correct.

I past upped nan ante by taking those briefing documents and feeding them to NotebookLM, which produced compelling, voice-annotated descent shows describing nan products. It was amazing, fascinating, and terrifying, each astatine nan aforesaid time.

Net-net-net though, since I'm my ain engineering and merchandise manager, nan 2 AIs together astir apt saved maine 60-80 hours of merchandise guidance work.

Analysis, SEO, and wide usage

Model: GPT-5.1 Auto

My go-to instrumentality for wide business support AI operations has been nan Plus tier of ChatGPT. While it tin occasionally beryllium incredibly uncooperative, it's besides mostly useful. Even if I didn't constitute astir AI for a living, I'd astir apt support my $20/month subscription. I usually support nan exemplary group connected Auto mode, which lets nan AI take nan processing level for each prompt.

My woman precocious bought a Plus subscription. She has been going awesome guns pinch ChatGPT, discussing nan nuances of belief arcana. She's some reasonably devout and reasonably academically inclined, truthful ChatGPT has been hugely nosy for her. I peculiarly bask nan moments erstwhile she calls ChatGPT retired successful a fib aliases fabrication, because that behaviour some amuses and annoys her astatine nan aforesaid time.

In position of applicable use, I find nan Plus tier to beryllium peculiarly adjuvant doing information analysis. I provender it large spreadsheets of different kinds of data, for illustration sentiment analysis, and person it crunch it, digest it, find nan cardinal issues, and springiness maine backmost results. I besides usage it to thief take SEO keywords for my YouTube videos.

Up until now, nary different chatbot has been arsenic broadly useful. However, Gemini 3 looks for illustration it mightiness springiness ChatGPT a tally for its money. Gemini 3 has only been retired for a week aliases so, which is why I don't person capable acquisition to comparison them. But, who knows, successful six months this class mightiness database Gemini 3 arsenic nan favourite exemplary alternatively of GPT-5.1.

Tools and models I don't usage

Among nan large names, I don't usage Perplexity, Copilot, aliases Grok. I cognize Perplexity besides uses GPT-5.1, but it's conscionable ne'er resonated pinch me. It's known for search, but nan fewer times I've tried immoderate searches, its results person been meh. Also, I can't guidelines nan truth that you person to log successful via email.

Copilot is an absorbing case. It did good successful nan free coding evaluations, but it's very Microsoft-centric. Despite nan truth that I was erstwhile all-Redmond, each nan time, I hardly ever usage Microsoft products now. I support my family subscription to Microsoft 365 Family because I cognize Excel and PowerPoint for illustration nan backmost of my hand. But I conscionable don't do Excel aliases PowerPoint activity overmuch anymore. Copilot is tightly integrated pinch those tools, truthful it conscionable hasn't proven to beryllium that compelling yet.

As for Grok, I was amazed really good it did successful my coding tests. But it doesn't clasp an agentic candle to either OpenAI's Codex aliases Claude Code. It besides did a awesome occupation connected my recreation itinerary trial successful my free chatbot tests. But it still seems inconsistent, and location are amended options easy available. I will proceed to trial it, and thing is formed successful concrete, but it's disconnected my database for now.

Also: Apple's iPhone App of nan Year is an AI instrumentality for group pinch ADHD - and it's free

You whitethorn besides person noticed that I didn't talk astir AI-generated video. I haven't really had overmuch of an juncture to do overmuch pinch AI video, different than my early Sora 2 evaluation for ZDNET. I'm judge I'll do more. When I do, I'll constitute astir it here. Also successful nan video category, I'm surely hoping that Apple adds much and amended AI capabilities to Final Cut Pro, which is my preferred video editor.

Oh yeah. Apple. Nope. Sorry. Nothing to spot here. Apple conscionable hasn't been connected my radar overmuch for AI. I tried its Coding Intelligence characteristic successful Xcode, and it conscionable collapsed each nan time. Apple's complete deficiency of AI practice successful articles for illustration this is bad news for nan company. The Cupertino elephantine really needs to up its AI game, and fast.

What astir you?

What astir you? Do you actively take different AI models depending connected nan task, aliases do you mostly instrumentality pinch 1 favourite tool? Have you recovered definite models to beryllium intelligibly amended for images, coding, research, aliases writing? Are location immoderate devices mentioned present that amazed you, aliases ones you powerfully disagree with? Let america cognize successful nan comments below.

You tin travel my day-to-day task updates connected societal media. Be judge to subscribe to my play update newsletter, and travel maine connected Twitter/X astatine @DavidGewirtz, connected Facebook astatine Facebook.com/DavidGewirtz, connected Instagram astatine Instagram.com/DavidGewirtz, connected Bluesky astatine @DavidGewirtz.com, and connected YouTube astatine YouTube.com/DavidGewirtzTV.