OpenAI Takes Aim at DeepSeek With New, Free ModelOpenAI Takes Aim at DeepSeek With New, Free Model

In tests, the two variants, o3-mini and o3-mini-high, fall short of challenging newcomer

Yiannis Antoniou, Head of data, analytics and AI, Lab49

February 3, 2025

3 Min Read
A small screen with the DeepSeek logo in front of a larger screen with the ChatGPT logo
Getty Images

As has become customary lately, OpenAI kept analysts, researchers and IT professionals busy this weekend. With the unveiling of its o3-mini release, OpenAI provided the first glimpse into the evolution of its reasoning models. This move came in response to DeepSeek’s r1 impact last week, aiming to showcase advances, particularly in coding, STEM and math. OpenAI introduced two variations of the model: o3-mini and o3-mini-high, with the latter allowing for additional "thinking" time to enhance responses. More notably, for the first time, OpenAI extended reasoning capabilities to free-tier users and significantly increased usage limits for paid users compared to its previous o1-generation models.

Performance and Comparisons

After extensive testing over the weekend, it’s evident that the model—especially at its high setting—has made noticeable improvements. It successfully solved complex coding problems that had previously stumped OpenAI’s earlier versions, resolved esoteric issues in existing code, and demonstrated clear advantages over DeepSeek in several areas.

However, DeepSeek still holds its own in terms of reasoning style. By revealing more of its thought process and maintaining a more human-like reasoning approach, it remains a compelling alternative. Despite the advancements in o3-mini, it still feels somewhat underwhelming – better than DeepSeek in raw performance but not a revolutionary leap. Given its pricing, nearly double that of its open-weight competitor, it raises the question: Would OpenAI have released o3-mini in its current form, with these generous usage caps and reduced pricing, had it not been feeling the pressure from DeepSeek?

Related:OpenAI Releases DeepSeek Challenger Model

A Fragmented Product Strategy

Beyond the model's performance, this release highlights an increasingly fragmented OpenAI product strategy. With no fewer than seven models now available - each with overlapping capabilities and missing features - it has become difficult for users to determine which model to choose. A clear capabilities matrix feels necessary at this point, as the differences between some models are nearly imperceptible. In contrast, DeepSeek offers a straightforward approach: Use DeepSeek r1. This simplicity is a competitive advantage, and unless OpenAI simplifies and streamlines its offerings, it risks losing market share. OpenAI has discussed plans to consolidate its model lineup, but until then, this disjointed strategy remains a concern.

The Bigger Picture

Overall, the o3-mini release proves that OpenAI is still in the fight - but this first post-DeepSeek round is, at best, a draw. Slightly improved capabilities at slightly worse price points, combined with a muddled product strategy, leave ample room for DeepSeek to refine and expand its own offerings. OpenAI seems to be aware of this and is likely shifting its focus toward its largest upcoming model, o3 (full version) - the release that the industry is eagerly anticipating. The real test will be whether it delivers a significant leap over DeepSeek and reestablishes OpenAI’s lead position. Unfortunately, we may have to wait for the Stargate initiative to materialize before seeing whether that generational leap happens only with software advances.

Related:Industry Weighs in on DeepSeek Launch

For now, OpenAI has taken a swing but hasn’t decisively blunted the DeepSeek challenge. However, this is still early days, and the race has many more laps to run.

About the Author

Yiannis Antoniou

Head of data, analytics and AI, Lab49, Lab49

Yiannis is head of data, analytics and AI at the specialist digital consultancy, Lab49. Yiannis has over 20 years of global experience in the financial, technology and consulting industries at major organizations including Goldman Sachs, JPMorgan, AIG, Pacific Global Advisors and Ernst & Young.

Sign Up for the Newsletter
The most up-to-date AI news and insights delivered right to your inbox!

You May Also Like