R1Tx?
#14
by
Blazgo
- opened
Why are you not releasing r1tx? is it on purpose?
Although R1Tx is the better model regarding benchmark scores, R1T2 is the model we liked to use much more. It's more fun to talk to and the shorter CoT pays out in every interaction with the model.
We also think that it is harder for users to chose the right model, if we release too many of them. Last not least: R1Tx is just one of many many models we evaluated internally and chose to not release.
Would you be able to release a private version to us, even if you don't release it. It's still faster than r1t2.
You can send us a DM, so that we can discuss it.
DM? on huggingface? not sure how to do that. Email?
The model card contains our contact addresses. Email or X, both are fine.
TNGHK
changed discussion status to
closed