Google unveiled the new version of the 2.5 Pro with superior performance in coding

Google new version of advanced language model 2.5 Peru Released in preview and claims that this model of competitors such as Deepseek R1 and Grok 3 Beta in the field Coding It works better.

The Jina 2.5 Pro’s 2.5 Pro was introduced in the first days of the new year and an improved version of it was available in May. A new version of the model is now released as a preview known as “Gemini 2.5 Pro Preview 06-05 Thinking”.

In its blog, Google mentions the new version as the smartest and most creative model ever released. The technology giant also plans to release a stable version in the coming weeks.

The new version of the 2.5 Pro is also available through platforms such as Google AI Studio and Vertex AI. According to Google, the new version of creative understanding, reasoning and response is better than previous versions. While Deepmind CEO Demis Hassbis (Google Artificial Intelligence Lab) at the Google I/O conference has introduced an improved version of Google’s coding model so far, the company now says version 06-05 is doing better.

The performance of the new version of the 2.5 Pro in benchmarks

Regarding the performance of this model in benchmarks, Google says that the new Jina 2.5 Pro has been able to improve in LMARNA and Webdevarena, respectively, respectively, respectively, and is currently ranking their ranking table. In Google’s assessments of competitors, this model has also surpassed other models, including Openai O4-MINI, Claude 4 Opus from anthropic, Grok 3 Beta and Deepseek R1.

Google also says that in designing the new version it has paid attention to previous feedback on Gemini 2.5 Pro; So that the structure of the answers is now better and the model is more creative. The new version is priced at $ 1.25 per million input token (non -cache) and $ 10 for the output.