Most models are going to require some heavy prompting to get them even close to the prose you’re looking for.

I’ve heard good things about Nous-Capybara-34B and Goliath 120b

Was wondering how long this would take to show up.

Do you plan to fine tune the 7b as well?

This style of captioning could be amazing for text to image datasets and i wouldn’t be surprised to see them take a jump in quality as well.

120b Goliath writing really is impressive. up there with cluade and gpt 4. Toppy also pretty good for a 7b

They have a premium 70b model that they have shown investors and enterprise customers.

They will likely opensource everything except the strongest model they have similar to what phind does.

I just hope we can still access the 70b model through the azure api.

We learned that merging models absolutely works and that the 34b yi model appears to be the real deal.

(Maybe we should merge some yi fine tunes in the future)

Intel has entered the game. Things are getting interesting.

If we ever get access to a mistral or yi 70b± model I think a lot of companies are going to be in trouble with their current models.

Can’t wait to see the benchmarks on these things.

Really want to see the numbers with mistral instead of llama 7b

This looks promising. I’d like to see a similar setup on the new mistral models if they ever release them.

Who is the wizard best defender? They look atrocious out there.

When do you expect to have benchmarks?