Last week, Google started calling all of its artificial intelligence products, including the Bard chatbot, “Gemini.” The name Gemini came from the language model that powers the company’s artificial intelligence tools. The tech giant made a brand new announcement today, showing that it is continuing its moves in the field of artificial intelligence without interruption.

The company announced the new version of Gemini in a blog post Gemini announced 1.5. Gemini 1.0 was released very recently, in December. Today’s announcement revealed that the new generation of the model has arrived just two months later.

Gemini 1.5 can process up to 1 hour of video, 11 hours of audio recording or 700,000 words simultaneously.

Gemini 1.5, the Language Model (LLM), offers serious improvements in many respects, even though it came very soon after its predecessor. Better understanding of the model, many a larger context window (the information it can process at once) and improved performance We will see. In fact, Google CEO Sundar Pichai says the Gemini 1.5 Pro, the first version of the new model, will deliver comparable performance to the Gemini 1.0 Ultra, the company’s current most advanced model. He adds that he did this with less processing power.

With Gemini 1.5, developers can develop much more functional models and applications. Google says Gemini 1.5’s context window will cap at 1 million tokens. This, 1 hour of video, up to 11 hours of audio, more than 30,000 lines of code or 700,000 words at once It means it can work. In a shared video, the model is shown a 44-minute film. Gemini 1.5 Pro can process information in a very short time and answer any questions about it.

For comparison: the context window of OpenAI’s latest GPT-4 Turbo model 128 thousand tokens. Gemini can offer almost 7.8 times as much. Let’s take into account that this amount is 32 thousand tokens in Gemini 1.0 Pro. So there is a huge improvement compared to the previous generation. The new model’s 1 million token context window will be available and tested to a limited number of developers. The company says it will comment on many issues, from latency to user experience.

Apart from this, the model managed to achieve 87% better results in the performance comparison with Gemini 1.0. According to the company, it also received perfect scores on the Needle In A Haystack (NIAH) assessment, which tests things like learning ability.

Gemini 1.5 is currently available to a limited number of developers and enterprise customers

*Gemini 1.5 Pro can understand and reason about intriguing details in 402 pages of text about Apollo 11’s journey to the moon.

Gemini 1.5 Pro, for now to a limited number of developers and business customers will be offered. It’s unclear when it will be widely released. However, we can say that it could eventually replace Gemini 1.0 Pro, which powers the chatbot “Gemini”, which is currently available for free to everyone. However, there is one part that we should not leave out.

Pricing will be offered based on the context window when the model is released to everyone. Gemini 1.5, worth 128 thousand tokens, comes standard. Versions that can hold up to 1 million tokens have different costs. So we probably won’t see the 1 million token context window in the open-to-all model.