Claude vs. Jurassic

Overview

Created in part by several of the developers of OpenAI's GPT-3, Claude is Anthropic's LLM trained with its research in developing helpful and safe AI systems.

Claude is Anthropic's flagship LLM and, as of April 2024, the most competitive cloud LLM offering alongside OpenAI's GPT series. Highly capable, Claude was also originally trained using Anthropic's "Constitutional AI" feedback method. Since the initial release, Anthropic has continued to improve the performance and cost of Claude, most notably in March 2024 with the release of the Claude 3 model family: Haiku, Sonnet, and Opus. As of April 2024, Haiku offers the most compelling cost-performance trade-off for non-critical applications, beating even GPT-3.5, while Opus has surpassed GPT-4 on the LMSYS leaderboard.

Initial release: 2023-03-14

Reference

https://www.anthropic.com/index/introducing-claude

Further Reading

• https://www.anthropic.com/index/constitutional-ai-harmlessness-from-ai-feedback
• https://www.anthropic.com/news/claude-3-family

Overview

Developed by AI21, a startup founded by luminaries from Stanford and Mobileye, Jurassic was one of the first LLMs to challenge GPT-3. While it includes some support for code, Jurassic is more focused on reading and writing use cases, perhaps linked to AI21's development of other tools such as Wordtune. Jurassic also introduced the MRKL system for interfacing with external, non-neural modules.

Initial release: 2021-08-11

Reference

https://www.ai21.com/studio

Further Reading

• https://www.ai21.com/blog/jurassic-x-crossing-the-neuro-symbolic-chasm-with-the-mrkl-system