gemini 1.5 pro , is a multilingual and multimodal ,a sparse mixture of expert (MOE) transformer based model , developed by gemini team .it was trained sometime after April 2023 (based on the knowledge cutoff) and released in may/2 2024 [1] ,google also claims that the model actually has a context length of 10M tokens (equivalent to approximately 5 days in audio , or 1440 pages of text or 10.5 hours of video) but only 2,097,152 tokens are available in the playground [2] .


training infrastructure gemini 1.5 was trained on multiple 4096-chip pods of google’s TPUv4 accelerators , distributed across multiple google datacenters ,and utilized JAX and ML pathways framework.[3]

Performance improvements compared to its predecessor gemini 1.0 pro [4]

gemini 1.5 pro is more efficient to train than 1.0 , its English query processing speed are over 650 characters per second.

Math, science and reasoning skills. (+49.6%) Multi lingual capabilities. (+21.4%) Coding skills. (+21.5%) Multimodal reasoning capabilities. (+31.5%) Video understanding capabilities. (+18.7%)

API Pricing related information [5]

2024/7/11

there are 2 types of API pricing

1-Free of charge. -Rate Limits 2 RPM (requests per minute) 32,000 TPM (tokens per minute) 50 RPD (requests per day) 2-Pay-as-you-go (prices in USD)

2-pay as you go. -Rate Limits 360 RPM (requests per minute) 4 million TPM (tokens per minute) 10,000 RPD (requests per day)

-Price (input) $3.50 / 1 million tokens (for prompts up to 128K tokens) $7.00 / 1 million tokens (for prompts longer than 128K)

-Context caching $0.875 / 1 million tokens (for prompts up to 128K tokens) $1.75 / 1 million tokens (for prompts longer than 128K) $4.50 / 1 million tokens per hour (storage)

-Price (output) $10.50 / 1 million tokens (for prompts up to 128K tokens) $21.00 / 1 million tokens (for prompts longer than 128K)