With Gemini 1.5, Google presented the next generation of its own AI language model shortly after renaming the chatbot Bard. The developers promise “dramatically improved performance”, with Gemini 1.5 Pro being at the same level of quality as Gemini 1.0 Ultra, but using less computing power. A highlight of Gemini 1.5 is the amount of tokens that the model can process, which is said to be one million.
In addition to words, Gemini 1.5 also processes videos, audio files and lines of code
According to Google, Gemini 1.5 is the AI language model that has the most extensive context window of all currently available models. In order to increase the efficiency of Gemini 1.5, the model is based on a new Mixture of Experts architecture (MoE).
By default, Gemini 1.5 currently works with 128,000 tokens, but the rollout of the context window will be gradual with one million tokens. Selected companies and developers can already use the full range of tokens via AI-Studio and Vertex AI as part of a private preview version.
According to Google, Gemini 1.5 Pro can currently process around an hour of video material, 11 hours of audio files, 30,000 lines of code or more than 700,000 words in one go. Google shows what this can look like in practice using the 402-page transcription of the Apollo 11 moon mission:
Recommended editorial contentHere you will find external content from [PLATTFORM]. To protect your personal data, external integrations will only be displayed if you confirm this by clicking on “Load all external content”:Load all external contentI agree that external content will be displayed to me. This means that personal data is transmitted to third-party platforms. Read more about our privacy policy.
If you want to try out Gemini 1.5 yourself, you’ll have to be patient. As already mentioned, the preview version of Gemini 1.5 Pro is currently only available to selected testers. A publication for a broader range of users will initially take place with a context window of 128,000 tokens. Prices for using Gemini 1.5 Pro have not yet been determined.
What: Google