Affiliate links on Android Authority may earn us a commission. Learn more.
The latest Gemini model is a coding and math whizz with 'parallel thinking'
May 20, 2025
- Google has announced Gemini 2.5 Pro Deep Think.
- This AI model is said to offer “incredible” performance in advanced math and coding tasks.
- The new AI model will be available to Gemini AI Ultra subscribers in the coming weeks.
Google periodically launches improved and specialized versions of its Gemini large language model. The company launched the Gemini 2.5 generation of models a couple of months ago, and it’s now introduced Gemini 2.5 Pro Deep Think.
The company used its I/O developer conference to reveal Deep Think, saying this experimental Gemini model offers “incredible” performance when it comes to advanced math and coding tasks. Google claims that Deep Think has an “impressive” score on the 2025 USAMO math benchmark, a leading score on the LiveCodeBench coding benchmark, and an 84% score on the MMMU multimodal reasoning benchmark.
Google says the model is able to consider multiple hypotheses and uses “parallel thinking” to deliver this improved performance.
The company explained that it’s taking time to conduct more safety checks and get more feedback from safety experts ahead of Deep Think’s release. It’ll also be using trusted testers to get more input before a wider launch.
Nevertheless, Gemini AI Ultra subscribers will get access to Gemini 2.5 Pro Deep Think when it eventually launches in the “coming weeks.” So you won’t have to wait too long to try out this new model.