Update Groq Models

[lumpidu]

The llama-3.1-70b-versatile model has been removed from the Groq API. With deepseek-r1's rising popularity, testing distilled versions on Groq's infrastructure could be worthwhile.

• add deepseek-r1-distill-llama-70B
• replace llama-3.1-70b-versatile with llama-3.3-70b-versatile

Note that Groq doesn't support the max. context lengths of 128K on the free tier. At least I get an API error if I exceed ~6000 tokens:

 Request too large for model deepseek-r1-distill-llama-70b in organization XYZ service tier on_demand
 on tokens per minute (TPM): Limit 6000, Requested 16955, please reduce your message size and try again.
 Visit https://console.groq.com/docs/rate-limits for more information.

[krschacht]

@lumpidu I think the issue is that you still had an assistant that was referencing the llama model you removed. I updated the assistants file.

Can you confirm this now works for you?