AI Lab · by Minh Nguyen · 2026-10-20
Fine-tuning vs prompting for domain tasks
With strong base models, prompting gets surprisingly far. When did fine-tuning actually pay off for you?
51 sparks · 18 comments
Comments
Grace Parker · 8 likes
Evals are getting saturated, we need harder ones
Marco Conti · 4 likes
Yeah same here
Cedar Otter · 0 likes
Not sure I agree but ok
Mina Chowdhury · 3 likes
Context window claims are always optimistic, test it yourself
Luca Bianchi · 14 likes
Ha, my wallet disagrees
Danylo Kovalenko · 8 likes
Well said
Narin Chai · 2 likes
The cost per token matters more than the leaderboard imo
Bilal Raza · 14 likes
Fair point
Camila Rocha · 1 likes
Open weights or it didnt happen
Amina Okafor · 8 likes
Haha my thoughts exactly
Curious Mango · 0 likes
My GPU just cried reading this
Curious Orbit · 14 likes
This ^
Nicolás Díaz · 14 likes
lol this made me laugh
Sleepy Coder · 0 likes
Tried it on my own dataset, results were mixed
Grace Parker · 2 likes
Not sure I agree but ok
Imogen Clarke · 0 likes
Prompt sensitivity is still the biggest pain in production
Antonia Silva · 3 likes
This ^
Nicolás Díaz · 0 likes
Depends but yeah mostly