Simply final week, Google rolled out Gemini 2.0 Flash to all Gemini app customers on cell and desktop. Constructing on that momentum, Google is now increasing entry and including to its Gemini mannequin choices.
Gemini 2.0 Professional is right here for Superior customers
On Wednesday, Google introduced it’s lastly releasing an experimental model of Gemini 2.0 Professional. The mannequin, which has Google’s largest context window in a mannequin of two million tokens, is the corporate’s most strong and superior mannequin for coding and sophisticated prompts.
The massive context window permits the mannequin to investigate and reference a strong quantity of knowledge directly, enhancing general efficiency and help by together with extra context. It additionally permits the mannequin to name on instruments resembling code execution, making it extra appropriate for quite a lot of duties.
Moreover, Gemini 2.0 Professional, the successor to 1.5 Professional unveiled a 12 months in the past, outperformed the remainder of Google’s Gemini fashions on a collection of benchmarks, together with the MMLU-Professional, which assessments for basic capabilities; GPQA (diamond), which assessments for reasoning; LiveCodeBench (v5), which assessments for code technology in Python; and MATH, which assessments for difficult math issues in algebra, geometry, pre-calculus, and extra.
This mannequin is offered as an experimental providing for Gemini Superior customers within the drop-down toggle. To be Gemini Superior person, you have to enroll within the Google One AI Premium plan that prices $20 per thirty days. In fact, Google could not overlook its developer person base, additionally providing it to builders by way of Google AI Studio and Vertex AI.
Introducing Gemini 2.0 Flash-Lite
Past that, Gemini 2.0 Flash, Google’s mannequin for quicker responses and stronger efficiency best for high-volume and high-frequency duties at scale, is changing into obtainable in additional Google merchandise, together with the Gemini API in Google AI Studio and Vertex AI, along with Gemini app entry, which was introduced final week.
Lastly, as a result of optimistic suggestions Google obtained on Gemini 2.0 Flash, it has now launched a brand new mannequin, 2.0 Flash-Lite. This mannequin retains the 1 million token context window, multimodal enter, velocity, and value of two.0 Flash, whereas providing higher high quality, in keeping with Google. This mannequin is offered in Google AI Studio and Vertex AI in public preview.
Google addressed security issues, reassuring the general public that the fashions have been constructed utilizing strategies designed to allow secure utilization, resembling new reinforcement studying strategies. The corporate additionally shared that the fashions went via automated red-teaming to evaluate safety dangers. This announcement follows Google’s publication of its Accountable AI: Our 2024 report, printed yesterday.