Major Release Google Multimodal Proprietary

Google Gemini 1.5 Pro

Google's first 1M-context model — multimodal needle-in-haystack champion

Released 2024-02-15 · ~175B · 1M tokens context · knowledge cutoff 2023-11

Overview

Google's next-generation model with breakthrough long context capabilities

Specifications

DeveloperGoogle
Release date2024-02-15
Model typeMultimodal
Parameters~175B
ArchitectureMoE Transformer
Context window1M tokens
Max output8192 tokens
Knowledge cutoff2023-11
LicenseProprietary
Input modalitiestext, image, audio, video
Output modalitiestext

Benchmarks

Benchmark Gemini 1.5 Pro Gemini ProΔ
MMLU 81.9% 79.1% +2.8
HumanEval 71.9% 67.7% +4.2
MATH 58.5% 32.6% +25.9

Availability

API Only Google AI Studio Vertex AI

What's new vs Gemini Pro

Direct comparison available in the model comparison tool →

Official references