Stop Overpaying for Simple Classifiers.
Use Sutro to dramatically reduce costs and speed up enterprise classification tasks. Create and reuse custom classification models that know your business inside and out.
import sutro as so
import polars as pl
from pydantic import BaseModel
new_products = pl.read_csv("march-2025-new-products.csv")
system_prompt = """
You are a helpful assistant that classifies products into categories:
Electronics, Home & Garden, Clothing, Toys & Games, Books, Movies & TV, Music, Art & Collectibles, Other
You will be shown a product's title and description. Return the category that best fits the product and nothing else.
"""
class NewProductClassification(BaseModel):
category: str
results = so.infer(
new_products,
column=["Title: " + "title", "Description: " + "description"],
system_prompt=system_prompt,
model="gemma-3-27b-it",
output_schema=NewProductClassification,
)
print(results.head())
Forget the training data
Skip the laborious process of gathering and paying for expensive training data. Use highly-capable, pre-trained models to save on upfront costs.
Iterate quickly to improve accuracy
Use Sutro’s fast prototyping capabilities to land on an accurate, versioned configuration within minutes.
Scale without the cost pains
Use your classifier on a few inputs, or a few million. Our throughput-optimized engine lets you run small, accurate classifiers for a fraction of the cost of other providers.
Go Deeper On Enterprise Classification
Hacker News is Obsessed with Aviation: Classifying 42 Million Posts with SLMs
An analysis of 40+ million Hacker News posts reveals that 0.62% are aviation-related, with the percentage steadily increasing over time. Using small language models, we classified 10.7B tokens of content to discover aviation's surprising popularity among technologists.
Workhorse LLMs: Why Open Source Models Dominate Closed Source for Batch Tasks
Open source LLMs now outperform closed alternatives at 90% lower cost for workhorse tasks. We analyzed benchmarks, pricing, and real-world performance to show exactly which models beat GPT and Claude—and by how much.
Document Categorization
Classify enterprise documents and emails for compliance, security, and hygiene.
Lead Scoring
Easily score inbound leads and prospects with a classifier suited to your business needs.
Churn Prediction
Continuously monitor customer behavior for signals of churn risk.
Sentiment Analysis
Easily monitor customer satisfaction from product reviews and support logs.
Customer Segmentation
Group users or prospects for better outreach targeting.
Fraud and Spam Detection
Identify bad actors, emails, resumes and more.
FAQ
70%
Lower Costs
1B+
Tokens Per Job
10X
Purpose-built classifiers for high-leverage teams.
Stop overpaying for simple classification tasks - switch to Sutro today.