The AI Arms Race

Photo by KaroGraphix Photography @karographixon Unsplash

Another day, another AI model. The latest salvos in the AI arms race are being fired, and OpenAI now has the o3-mini and the deep research agent.

The backdrop to this is the pressure to keep pace with DeepSeek, the Chinese AI company making leaps in AI (although OpenAI claim DeepSeek may have pinched its data, but that’s a story for another day).

Is this arms race the beginning of a larger AI war, or is this a case of tech companies trying to keep up appearances at the dawn of a new computing age?

The Battle for AI Supremacy

OpenAI has sat at the peak of inflated expectations for AI, rolling out sequential ChatGPT updates and creating an ever-expanding array of subscription tiers for their engaged audience.

Then along came DeepSeek, leading an explosion of competitors in the field that have forced OpenAI to shift gears and improve their models.

o3-mini is being framed as a response to competition, particularly in STEM, coding, and reasoning capabilities. On paper, it sounds like a bold strategic move, but when you look under the hood these incremental updates seem to be more like marketing maneuvers than actual innovation. They seem to be adding features that look good on a marketing website, but lack the punch users require from a cutting-edge AI system.

Smaller, Faster, Smarter…Cheaper?

The key selling point of o3-mini is efficiency. OpenAI claims that this model “advances the boundaries” of small, cost-effective AI reasoning.

It seems to me that they are running scared of the cheaper DeepSeek models and are playing catchup. Nobody is going to use deep research and confuse it with AGI, no skeptic is going to think that hallucinations are a thing of the past or that guardrails are in place to really make this tech safe for users. OpenAI is framing releases as wins for users, but the actual victory is simply OpenAI’s bottom line.

Incremental tweaks will not make AI trustworthy, and the application of AI is frankly too limited at the moment to make a real change to the world. We’re still at the beginning, and commoditizing AI isn’t the solution to the complex problems we see in 2025.

The Lock-In

OpenAI is keen to get free users on the platform, presumably with hopes that they can be monetized sooner rather than later. So, it makes sense that OpenAI lets users access the reasoning model (by using the friendly-looking “reason” option), which does indeed feel generous move. Yet these same users are the ones who are least likely to prevent their data being used to train models and prevent themselves becoming the product.

I’ve noticed at work we seem stuck on the offerings from OpenAI. Indirectly we use Copilot due to the availability of plugins in Visual Studio Code, and we use OpenAI’s API to ensure access to ChatGPT is within our own policy rules. Any question of moving from OpenAI comes with the reply that we are already invested in the ecosystem, and if this is my experience how many more developers experience the same?

Giving subscribers higher message limits goes some way to providing more value, but it is also another move that keeps paying users locked in and unable to bear the cost of changing platform.

The One-Sided Arms Race

If this is an arms race, it’s a one-sided one. OpenAI still holds the technological and market advantage, and DeepSeek, while promising, is more of a distant challenger than a direct threat. The bigger question is whether these updates are truly revolutionary or just another attempt to keep the AI hype train rolling while companies figure out how to squeeze more money out of it.

Conclusion

Whether it’s OpenAI, DeepSeek, or the next AI upstart, the one thing that seems certain is that users will still gain incremental improvements to current models, wrapped in increasingly dramatic press releases.

You know what. This is an AI war. But the real battle isn’t OpenAI vs. DeepSeek. It’s marketing vs. reality.