Learn
Practical guides to fine-tuning, distillation, and deploying small language models.
Accuracy, F1 or Exact Match: Which Metric Should You Use?
A side-by-side comparison of accuracy, F1, exact match, ROUGE-L, LLM-as-a-judge and the tool-calling metrics, with the task each one fits and the specific way each one lies to you.
Agent Frameworks vs a Fine-Tuned Tool-Calling Model
A framework wires tools to a model and manages state; it doesn't make the model pick the right tool. Which layer your failures actually live in, and why most systems need both.
Alternatives to the OpenAI Fine-Tuning API
The four categories of alternative, compared on where the model runs and what you own afterwards: self-hosted training libraries, managed open-weight platforms, cloud provider tuning services, and distillation platforms.
Can a Small Model Beat Its Teacher?
Yes, on narrow tasks. A 0.6B student beat a 120B teacher by 29 points and a 350M student beat the same teacher on two of three tool-calling benchmarks. The measured figures and why it happens.
Cost Per Million Tokens: Local vs API
A side-by-side of self-hosted small-model rates against per-token API pricing, and why the two figures aren't directly comparable even when both are quoted per million tokens.
Distillation vs Pruning
Pruning deletes parts of a trained network and keeps what survives; distillation trains a fresh small model on a large one's behaviour. Why published pruning pipelines end with a distillation step anyway.
Does Base Model Accuracy Predict Fine-Tuned Performance?
Across four mid-size models on seven tasks the base ranking inverted after fine-tuning: the worst zero-shot model tied for best once trained. Here are the ranks, the method, and what still transfers.
GPU Requirements and Cost for SLM Inference
Weight-size arithmetic for the distil labs student catalog alongside one measured serving benchmark (222 RPS and 7.6 GiB for a 4B model) and how a GPU-hour bill becomes a per-request price.
How Big Should Your Test Set Be?
Published distil labs test sets run from 10 to 374 examples, with 50–250 the working band. Here are the real sizes, the platform defaults, and the resolution each size actually buys you.