I quit ChatGPT for a free, private, and local AI called Ollama - here's why

Jack Wallen/ZDNET

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

ZDNET's cardinal takeaways

• Ollama is simply a user-friendly, locally installed AI.
• You volition beryllium amazed astatine the benefits you summation from this app.
• Not lone is Ollama escaped and private, but it's besides open-source.

Ollama is simply a tiny subordinate successful the AI game, but it should beryllium overmuch bigger than it is. This installable AI has respective benefits you won't find with the likes of ChatGPT, and those benefits are what support maine from ever utilizing a much accepted service.

I've written astir Ollama rather often implicit the past year, but I thought it was clip to stock the reasons I take to usage this app implicit immoderate other.

Also: Want section vibe coding? This AI stack mightiness regenerate Claude Code and Codex - for free

(Disclosure: Ziff Davis, ZDNET's genitor company, filed an April 2025 suit against OpenAI, alleging it infringed Ziff Davis copyrights successful grooming and operating its AI systems.)

What is Ollama?

Before I get into the why, let's speech astir the what. 

Ollama is simply a free, open-source instrumentality that you tin download and instal connected your machine (Linux, MacOS, oregon Windows). With Ollama, you tin tally ample connection models (LLMs) straight connected your system. The main contented to see is your computer's speed. AI requires a just spot of horsepower to beryllium useful. That doesn't mean you can't usage it connected a midrange machine, but if you're utilizing it connected mid- to low-end hardware, it'll tally much slowly, and you mightiness not beryllium capable to multitask portion it churns done your queries.

Also: OpenAI's caller representation watermarks marque it easier to spot AI fakes - here's how

However, if your instrumentality has an Nvidia GPU, it volition tally faster (even with midrange hardware) due to the fact that it'll offload a batch of the processes onto the GPU, leaving the CPU to bash different things. To tally Ollama smoothly, your strategy should person a minimum of:

• CPU: Any modern processor
• System RAM: 16 GB
• GPU (Recommended): Nvidia with 8GB+ VRAM oregon an Apple Silicon Mac (M1/M2/M3) with 16GB+ unified representation for automatic GPU acceleration

Ollama offers a user-friendly GUI for some MacOS and Windows, but tin besides beryllium utilized from the bid line. On Linux, you tin usage the Ollama CLI, but determination are besides GUIs (such arsenic Alpaca and Msty) that tin work.

With Ollama, you tin download galore antithetic LLMs. There's a vast library of models that tin beryllium downloaded -- you'll find DeepSeek, Gemma, Qwen, Mistral, Gpt-OSS, Llama, and galore others.

Now that you recognize what Ollama is, let's speech astir why you should beryllium utilizing it.

1. It's free

Ollama is free. Since I've been utilizing it, I've not paid a azygous penny for the app oregon the models. Just instal Ollama, propulsion a model, and go. You don't person to wage for the app, the models, oregon usage. 

It's free. Totally and everlastingly free.

It's besides open-source, so… bonus.

Also: Google's AI features conscionable got much confusing

2. It's private

This is simply a large 1 for me. I won't usage nationalist oregon for-profit AI due to the fact that I cognize my queries and the responses tin beryllium utilized by 3rd parties. The companies down for-profit AI tin cod your accusation and usage it for assorted purposes, and that's not thing I americium OK with. It's not that I query with immoderate idiosyncratic accusation (I mostly usage AI for research), but I inactive don't privation those queries collected and utilized to make a illustration of me. 

Also: The champion unafraid browsers for privacy: Expert tested

Just like my browsers, I privation my AI to beryllium arsenic backstage arsenic possible, and Ollama is truly the lone mode to marque that happen.

3. It doesn't situation the electrical grid

You've heard of however AI uses an inordinate magnitude of energy, right? A study this period from the International Data Center Authority (ICDA) notes that an estimated 6% of full US energy usage goes to information centers. "The US is by acold the world's largest information halfway location, with 43% of planetary consumption. Data centers devour 29.2GW of energy successful the US, and present devour 6 % of the nation's electricity."

This UN report details the biology concerns surrounding large-scale AI deployments. "Most AI servers are stored successful information centers, which nutrient physics discarded and tin incorporate toxic chemicals, specified arsenic mercury and lead. Data centers besides necessitate ample amounts of h2o for operation and to chill the electrical components. "
Also: Worried astir AI's soaring vigor needs? Avoiding chatbots won't assistance - but 3 things could

By utilizing locally-installed Ollama, I debar adding to that problem. I could instrumentality that 1 measurement further and instal Ollama connected a laptop, truthful I'm lone utilizing artillery power. 

Ollama is environmentally affable and does not enactment immoderate strain connected the electrical grid. 

4. It's much flexible

When you're utilizing the likes of ChatGPT, you're constricted to the LLMs you tin take from. With Ollama, you person a monolithic room of LLMs astatine your fingertips. You tin adjacent instal aggregate models and take which 1 to usage for each query. 

Also: How I easy added AI to my favourite Microsoft Office alternative

That's a flexibility you won't find successful astir for-profit AI services.

5. It's LAN-able

I person 1 lawsuit of Ollama installed connected a server wrong my LAN. With that up and running, I tin either usage a web-based interface oregon link a section GUI app to the server, truthful I'm not utilizing CPU/GPU connected my laptop oregon desktop. That's a large enactment due to the fact that offloading the AI process to a server means I don't person to interest astir queries bringing little almighty hardware to a halt. On apical of that, I lone person to instal Ollama connected 1 instrumentality to usage it connected each instrumentality connected to my LAN.

6. Offline accessible

Let's accidental you suffer your net connectivity, oregon you're retired successful the mediate of obscurity without a awesome oregon Wi-Fi, and you request (for immoderate reason) to usage AI. If you usage a third-party, cloud-based AI, you're retired of luck. If you person Ollama installed connected your laptop, you're precise overmuch successful luck, due to the fact that you don't person to person a web transportation to usage it. That besides means you tin tally Ollama connected an air-gapped instrumentality for adjacent much privacy.

Also: How I provender my files to a section AI for better, much applicable responses

It's win-win each around.

I would highly urge you effort a locally installed lawsuit of Ollama. It's overmuch easier to get up and moving than you deliberation (similar to installing immoderate app connected MacOS oregon Windows), it's cheap, flexible, secure, and environmentally friendly. 

Who doesn't privation each of that?