---
title: "Toolformer: Language Models Can Teach Themselves to Use Tools"
description: "Toolformer asks how a language model can learn when and how to call external tools without a large hand-labeled tool-use dataset. Read it to see how the model proposes calls inside ordinary text, executes them, and keeps only those whose returned result lowers continuation loss enough to become a training example."
canonical_url: "https://fanout.sh/daily/2026-08-20-toolformer"
md_url: "https://fanout.sh/daily/2026-08-20-toolformer.md"
last_updated: "2026-08-20"
access: "public"
---

# Toolformer: Language Models Can Teach Themselves to Use Tools

Toolformer asks how a language model can learn when and how to call external tools without a large hand-labeled tool-use dataset. Read it to see how the model proposes calls inside ordinary text, executes them, and keeps only those whose returned result lowers continuation loss enough to become a training example.

## Paper details

- Authors: Schick et al.

- Venue: NeurIPS 2023

- Track: AI Research

- Difficulty: Easy

- Reading time: 40 min

- Original paper: https://proceedings.neurips.cc/paper_files/paper/2023/hash/d842425e4bf79ba039352da0f658a906-Abstract-Conference.html

## What you will learn

- How a few demonstrations turn the base language model into a permissive annotator that proposes API-call positions and arguments.

- Why Toolformer compares future-token loss with and without each executed result, and what that threshold does not certify.

- How retained calls become fine-tuning data, why sufficiently capable base models benefit most, and where single-call learning stops short of a complete agent.

## Continue exploring

[Explore tool use and function calling](https://fanout.sh/ai/glossary#tool-use-function-calling): Continue into Fanout's tool-use glossary and connect model-generated calls to external executors, structured arguments, permissions, validation, latency, and modern agent harnesses.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md