[Feature Request]: Add dated Mistral AI models

[GICodeWarrior]

Feature Description

Any reason not to add the versioned MistralAI models? Also, many more of their models claim to support function calling than what llama_index has listed. Any reason not to add those as well?

Completion models (w/ context size):

% curl -s --header "Authorization: Bearer $MISTRAL_API_KEY" https://api.mistral.ai/v1/models | jq -r '.data.[] | select(.capabilities.completion_chat) | [.id, .max_context_length] | @csv'
"ministral-3b-2410",131072
"ministral-3b-latest",131072
"ministral-8b-2410",131072
"ministral-8b-latest",131072
"open-mistral-7b",32768
"mistral-tiny",32768
"mistral-tiny-2312",32768
"open-mistral-nemo",131072
"open-mistral-nemo-2407",131072
"mistral-tiny-2407",131072
"mistral-tiny-latest",131072
"open-mixtral-8x7b",32768
"mistral-small",32768
"mistral-small-2312",32768
"open-mixtral-8x22b",65536
"open-mixtral-8x22b-2404",65536
"mistral-small-2402",32768
"mistral-small-2409",32768
"mistral-small-latest",32768
"mistral-medium-2312",32768
"mistral-medium",32768
"mistral-medium-latest",32768
"mistral-large-2402",32768
"mistral-large-2407",131072
"mistral-large-2411",131072
"mistral-large-latest",131072
"pixtral-large-2411",131072
"pixtral-large-latest",131072
"codestral-2405",32768
"codestral-latest",32768
"codestral-mamba-2407",262144
"open-codestral-mamba",262144
"codestral-mamba-latest",262144
"pixtral-12b-2409",131072
"pixtral-12b",131072
"pixtral-12b-latest",131072

Function calling models:

% curl -s --header "Authorization: Bearer $MISTRAL_API_KEY" https://api.mistral.ai/v1/models | jq '.data.[] | select(.capabilities.function_calling) | .id'
"ministral-3b-2410"
"ministral-3b-latest"
"ministral-8b-2410"
"ministral-8b-latest"
"open-mistral-7b"
"mistral-tiny"
"mistral-tiny-2312"
"open-mistral-nemo"
"open-mistral-nemo-2407"
"mistral-tiny-2407"
"mistral-tiny-latest"
"open-mixtral-8x7b"
"mistral-small"
"mistral-small-2312"
"open-mixtral-8x22b"
"open-mixtral-8x22b-2404"
"mistral-small-2402"
"mistral-small-2409"
"mistral-small-latest"
"mistral-medium-2312"
"mistral-medium"
"mistral-medium-latest"
"mistral-large-2402"
"mistral-large-2407"
"mistral-large-2411"
"mistral-large-latest"
"pixtral-large-2411"
"pixtral-large-latest"
"codestral-2405"
"codestral-latest"
"codestral-mamba-2407"
"open-codestral-mamba"
"codestral-mamba-latest"
"pixtral-12b-2409"
"pixtral-12b"
"pixtral-12b-latest"

Reason

No response

Value of Feature

No response

[logan-markewich]

Feel free to add them! I haven't been keeping on top of it (and also you are the first to ask!)

[logan-markewich]

In fact, if you can wire in that API call instead of hardcoding, that would be even better

[GICodeWarrior]

If I implement the API call, should it be a one-time call (cached in memory), or should it be called every time a context size is checked, etc.?

Alternately I was considering sticking a script next to the README that outputs the model info to paste into the actual code.

[logan-markewich]

probably caching it for the lifetime of the class instance makes sense 👀