Gemini (FKA Bard) vs. Mistral Large

Overview

Gemini is Google's LLM chat product, based on its Bard model, which was based on its Pathways Language Model (PaLM) which succeeded LaMDA. Originally released with limited capabilities, Bard became increasingly capable in the second half of 2023. Bard built upon Google's research in the transformer architecture, Gopher scaling, Chinchilla compute-optimal scaling, and Sparrow dialogue finetuning. As the successor to Bard, Gemini achieves GPT-4-level performance along with the ability to write code and process images.

Initial release: 2023-03-21

Reference

https://bard.google.com

Further Reading

• https://ai.googleblog.com/2022/04/pathways-language-model-palm-scaling-to.html
• https://www.deepmind.com/publications/an-empirical-analysis-of-compute-optimal-large-language-model-training
• https://www.deepmind.com/blog/building-safer-dialogue-agents
• https://blog.google/technology/ai/google-gemini-ai/#availability

Overview

Mistral Large is Mistral AI's flagship proprietary model, competitive with GPT-4 class models. It excels at complex reasoning, multilingual tasks, and code generation, and is available via the La Plateforme API and major cloud providers.

Released in February 2024, Mistral Large is Mistral AI's top-tier commercial offering. It supports a 32K token context window and achieves top-three ranking on MMLU at time of release, behind only GPT-4 and Google's Gemini Ultra. Mistral Large 2 followed in July 2024 with a 128K context window and improved instruction following. Both versions are available via Mistral's own API and through Azure and AWS Bedrock.

Initial release: 2024-02-26

Reference

https://mistral.ai/news/mistral-large/

Further Reading

• https://mistral.ai/news/mistral-large/
• https://mistral.ai/news/mistral-large-2407/