Google releases its own ‘reasoning’ AI model

Google introduces Gemini 2.0, an experimental reasoning AI model in its AI Studio.

: Google unveiled the Gemini 2.0 Flash Thinking Experimental model, aimed at multimodal reasoning in AI Studio. Supposedly, this model can tackle complex problems across programming, math, and physics with enhanced reasoning capabilities. However, initial tests show it still faces challenges, such as accurately counting letters in words. With an increase in reasoning model developments, this area continues to experience rapid growth.

Google has launched Gemini 2.0 Flash Thinking Experimental, a new reasoning AI model, in its AI Studio. This model claims to excel in multimodal understanding, reasoning over complex issues in areas like programming and physics.

Logan Kilpatrick mentioned it as the beginning of Google's reasoning journey, while Jeff Dean emphasized improved reasoning from extended inference computation. Despite its promise, the model still struggles with seemingly simple tasks, highlighting the experimental nature of its current iteration.

The emergence of reasoning models is gaining traction, with companies like DeepSeek and Alibaba entering the field. Skepticism remains about the sustainability and costs associated with these models, as shown by wider investment in their development by firms like Google.

AIReasoning ModelsGoogle