Google has recently revealed the debut of Gemini 1.0. An advanced large language model that claims superiority over OpenAI GPT-4 in various aspects. In June, Google CEO Pichai introduced Gemini at the I/O Developers Conference, proclaiming it as the most potent and versatile language model to date.
In a high-profile unveiling, Google’s latest creation, the Gemini 1.0 large language model, takes the spotlight, showcasing its dominance over OpenAI GPT-4. Gemini’s “state-of-the-art performance” on multiple benchmarks outshines GPT-4, marking a significant milestone. To cater to diverse scenarios, Google has introduced three distinct versions of Gemini: Gemini Ultra, Gemini Pro, and Gemini Nano.
- Gemini Ultra, the most robust and extensive model, excels in handling highly intricate tasks.
- Gemini Pro stands out as the optimal model for scaling a variety of tasks.
- Gemini Nano, known for its efficiency, is tailored for devices like mobile phones.
Scheduled for release on December 13, Gemini Pro’s Gemini API will be accessible to developers and enterprise customers through Google AI Studio or Google Cloud Vertex AI. Meanwhile, Gemini Ultra is exclusively available for invited customers, developers, partners, and security experts for early experimentation and feedback. Google plans to extend access to more developers and enterprise customers in the early part of next year.
Gemini Ultra showcases “advanced results” surpassing the current era across 30 out of 32 widely adopted academic benchmarks. Notably, on the Massive Multi-Task Language Understanding (MMMU) benchmark, it achieves a score of 90.0%, marking the first instance of a model surpassing human experts. Furthermore, Gemini Ultra leads in the new MMMU benchmark, achieving an impressive score of 59.4%.
Read Also: Google And European Telecom Giants Call on EU to Force Apple to Open Up iMessage
