https://www.boundaryml.com logo
I'm curious if somebody tried any local
# general
r
I'm curious if somebody tried any local model (or any other models). to write baml code. I burned all my Claude/Codex subscriptions these days πŸ˜‚
v
i haven't yet but unsure how the models would do! i suspect you likely need a sonnet+ model to make it work, but i wonder if deepseek can do it well given how good they are now
x
I'm just about to run some evals on local models writing effectively "BAML lambdas" or is it a BAML lamb-da? πŸ‘ I don't have the hardware to run anything large, but instead of expecting repo-scale, I wanted to approach it from a "swarm of micro-functions" orchestrated by a frontier model that doesn't write any code.
r
I was thinking to test the new qwen3.8 27b, but I'm away this week and I can’t host it
a
am testing 2 sparks with qwen 3.8 right now, will add baml evals to the list πŸ™
2 Views