/Tech6h ago

Rohit Krishnan proposes mathematical model to calculate the financial breakeven point of fine-tuning versus frontier APIs

The formula assumes six months of development yields 90% savings.

2330217.6K

Original post

rohit@krishnanrohit#1319inTech

This is very interesting. What it really shows is an equation, eg around 6 months of a frontier-level team can build something which reduces inference by 10x per task.

Or N = F/(0.9*C), N is # tasks, F is fixed cost, C is frontier inference $. You can calculate your own breakeven.

Mira Murati@miramurati

Bridgewater used their unique financial knowledge and partnered with us on @tinkerapi to fine-tune a model that helps their analysts focus on what's important. Experts improving AI that empowers experts. https://thinkingmachines.ai/news/learning-to-replicate-expert-judgment-in-financial-tasks/

2:44 PM · Jul 1, 2026 · 6.3K Views

Sentiment

Sentiment building, check back later.

Cluster Engagement

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

THINKING MACHINES LABVia

#46

Posts from X

Most Activity

VIEWS1.2KBOOKMARKS1LIKES2REPLIES1

rohit@krishnanrohit

On Bridgewater training models to help replace frontier model use

rohit@krishnanrohit

This is very interesting. What it really shows is an equation, eg around 6 months of a frontier-level team can build something which reduces inference by 10x per task.

Or N = F/(0.9*C), N is # tasks, F is fixed cost, C is frontier inference $. You can calculate your own breakeven.

5h1.2K21