After residing solely with the Gemini 2.0 Flash compact version since December 2024, Google has finally unveiled the larger model, Gemini 2.0 Pro, which is still in the Experimental status for testing via Google AI Studio and Vertex AI.
Google claims that the Gemini 2.0 Pro excels in coding capabilities, handling complex prompts, demonstrating a deep understanding of the world, and reaching the highest reasoning level among all Gemini models. It also supports a large context window of 2 million tokens, equivalent to the previous Gemini 1.5 Pro model.
The benchmark comparison brought by Google between Gemini 2.0 Flash and Gemini 2.0 Pro showcases improvements on almost all fronts, although there are slight areas where Gemini 2.0 Flash still outperforms, such as in FACTS Grounding. It is anticipated that Google will gradually update the Gemini 2.0 Pro with incremental sub-versions before its official release, mirroring the process followed with Gemini 2.0 Flash.
In addition to the flagship model, Google has introduced Gemini 2.0 Flash-Lite, a downsized version of Gemini 2.0 Flash aimed at replacing Gemini 1.5 Flash by providing better responses while maintaining speed-price parity with the predecessor (refer to pricing table for details).
Meanwhile, the standard version of Gemini 2.0 Flash is now in the generally available status, allowing access through API in Google AI Studio and Vertex AI.
For clarity, the Gemini 2.0 model family currently comprises four sub-models: Gemini 2.0 Flash, Gemini 2.0 Pro (Experimental), Gemini 2.0 Flash-Lite, and Gemini 2.0 Flash Thinking (Experimental).
Source: Google, Google for Developers
TLDR: Google expands its Gemini AI model lineup with the experimental Gemini 2.0 Pro, highlighting advanced coding capabilities and high reasoning skills, alongside releasing the Flash-Lite version for improved responses and maintaining cost-efficiency. Gemini 2.0 Flash now offers broad availability through API access in Google AI Studio and Vertex AI.
Leave a Comment