Google has managed to launch its Gemini 1.5 Pro and Flash models smoothing the path for the developers by ensuring better performance, and reduced costs. The search engine giant also introduced stable releases of the Gemini 1.5 API, which present advances in several factors including code generation, mathematics, reasoning, and video analytics.
Announcement on September 24, the Gemini 1.5 Pro (gemini-1.5-pro-002) and Gemini 1.5 Flash(gemini-1.5-flash-002) are vast improvements over the earlier models. Google claims the maximum limit on the new models is three times higher and latency cuts down and the Pro model is now over 50% cheaper.
Google Improves Model Factuality Accuracy
Google Found out, that the two models offer improvements in factuality, hallucinations in the model were lowered, and better instruction following was achieved. They also provided improved understanding of 102 languages, SQL generation, and comprehension of audio and documents in addition to better multilingual comprehension.
Google also cut the summarization length and provided advice to the chat product developers on suggestions to make API conversations better. Effective October 1, 2023, the Gemini 1.5 Pro API will make lower the price of its services to the consumers who are willing to pay less than 128,000 tokens, with input tokens going for 64%, while the output tokens going for 52% less than before.
Google is also increasing the structure growth for the paid tier of the 1.5 Flash to 2000 RPM and for 1.5 Pro to 1000 RPM from the previous limits of 1000 and 360 respectively.
Concurrently, Google has introduced its new model Gemini 1.5 Flash-8B, which is a more compact model and, moreover, is of lower illegal benchmarks but demonstrates better results within toneless and multimodal tasks performed in writing. All versions are already presented at Google AI Studio and via Gemini API.
At the same time, another company Open AI has started to implement ChatGPT with an “Advanced Voice” feature. This new mode allows to shift from typing to verbal interaction with the AI, incorporating the voices of five new characters: Arbor, Maple, SXol, Spruce, and Vale. To these are added the existing voice options of Breeze, Juniper, Cove, and Ember.
