Did you know Google Rolls Out New Gemini 2.0 Flash Thinking Model With Advanced Reasoning

 

Tech giant Google has silently rolled out its Gemini 2.0 Flash Thinking Model with advanced reasoning capabilities.

The company shared the major update for its Gemini model that now gives users a clear reasoning process as part of the explanation. Similarly, it reaches new highs in terms of performance records for tasks in mathematics and science. Lastly, it provides a free-of-cost option with data limits when compared to the premium services offered by OpenAI.

The matter has certainly made the makers of ChatGPT think twice on this front including how the competition is rising in the AI world. The latest offer from Google AI Studio was able to attain a 73.3% score on AIME and a 74% score for setting a science benchmark on GPQA. it’s clearly more advanced in terms of reasoning and these performance records are proof.

The CEO of Google’s DeepMind shared more on the release including how the company has been working long and hard for these systems. It’s exciting for them to see the ideas brought to life in the form of reliable foundation models.

One of the model’s most striking offerings is the fast processing of one million text tokens. This is almost a fivefold rise from that seen on OpenAI’s o1 Pro Model. The new world record also ensures faster times for giving out accurate responses. During this new and expanded window for context, the model can analyze several studies at the same time. Therefore, it provides researchers with the ability to work alongside huge data volumes.

The timing for the release is certainly very critical. It was a moment when OpenAI shared its breakthrough o3 model that got an 87.7% GPQA Diamond score. However, the fact that Google is giving users the model for free during this beta testing stage says so much. Yes, there are limits for usage but it might end up being the most popular option for developers and companies looking for alternatives to the costly $200 per month subscription from OpenAI.

Other improvements worth mentioning include higher reliability and fewer contradictions between the thoughts generated by the model and the final output produced. There’s also a new native code capability installed for execution. This gives developers the chance to run and experiment directly inside the system. There are more safeguards as well so we do see it become a major contender for research purposes and commercial apps. 

 Image: Google