No More Excuses: OpenAI’s Thinking Models Go Open Source

A few days ago, I decided to put OpenAI's new open-source "thinking" models to the test.

It took me less than 20 minutes.
I downloaded the model through Ollama, wrote a small script (ironically, with the help of another AI), and ran it locally on my MacBook Pro.

The result? The gpt-oss:20b model was up and running, fast, and entirely under my control—no data leaving my laptop.

For years, businesses—especially in regulated industries—have raised the same objection:

“We can’t use AI because we can’t risk sending sensitive data to the cloud.”

That excuse is now gone.

What’s Changed

Two weeks ago OpenAI has made two reasoning models openly available under an open-source license. That means:

• You can download them, run them locally, and keep data inside your infrastructure.
• Security and compliance barriers are dramatically lowered.
• Experimentation is no longer reserved for companies willing to send data outside their walls.

The Catch

Of course, open source doesn’t mean “full package.”

• These models work, but they don’t come with the tooling layer that OpenAI’s hosted services provide (like web search, memory, analytics dashboards, or integrations).
• You’re responsible for performance tuning, deployment, and scaling.

So the trade-off is simple:
local = security and flexibility, but fewer features.

Why SMBs Should Pay Attention

For SMBs, this is a breakthrough moment:

• No need for lengthy vendor reviews or legal approvals.
• No subscription fees—just your own hardware.
• A chance to move fast, experiment safely, and build solutions that fit your business without worrying about data leaving your control.

This levels the playing field. A small firm can now harness cutting-edge AI with nearly the same raw power as a Fortune 500 company.

The Enterprise View

For Enterprises, the story is more complex.

Large organizations thrive on integration, compliance, and scale. An open-source model sitting on a server may not check the boxes for monitoring, governance, or multi-departmental use.

Compared to OpenAI’s hosted platform, these models will feel bare-bones. Enterprises will need to invest significantly in building the missing pieces before they see real productivity gains.

The Bigger Question

Here’s the real shift: Data security is no longer a blocker.

The question for every business leader becomes:

• If you’re an SMB: How quickly can you start experimenting safely?
• If you’re an Enterprise: Do you invest in building your own AI stack, or leverage richer hosted services?

Either way, the ground has moved. AI adoption is no longer about whether it’s safe. It’s about how you’ll use it to move your business forward.

What are your thoughts on this topic? Reply to our newsletter or connect with us on LinkedIn.