Anthropic have released the Claude 3 family: Announcement.

Early comments from my filter bubble:

• not available in the EU
• hosting options also include Microsoft Azure, in addition to AWS?
• overreaching guardrails: Claude still refuses work at times, e.g. coding a website may be gaming and misrepresenting benchmark scores. The comparisons they have published are against the original GPT-4, not the current ones
  • https://x.com/alexalbert__/status/1764722513014329620?s=20
  • https://twitter.com/abacaj/status/1764752690360238220
• real-world task performance seems mixed, some praise coding abilities, others don’t https://twitter.com/rishdotblog/status/1764720887331270754
• biggest model (GPT-4 competitor “Opus”) more costly than GPT-4 (Turbo?)
• vision on par with GPT-4V?

(I haven’t tried any of this myself. Let me know if you’d like my Amz Chatbot to be updated soon’ish.)

Update 1: There seems to be consensus that it is better than 2.1, especially for Summarization.

Update 2: Chat interface for Claude 3: https://huggingface.co/spaces/ndurner/claude_chat

This requires an API key from Anthropic themselves as the biggest model “Opus” in particular is not available through AWS.