# microsoft/ToRA

ToRA is a series of Tool-integrated Reasoning LLM Agents designed to solve challenging mathematical reasoning problems by interacting with tools [ICLR'24].

Repository: https://github.com/microsoft/ToRA
Canonical: https://ross.abutalabs.com/products/microsoft-tora
Homepage: https://microsoft.github.io/ToRA/
Language: Python
License: MIT
License Family: permissive
Topics: autonomous-agents, language-model, llm, mathematical-reasoning, tool-learning
Last push: 2024-02-22T09:46:48+00:00

## Health v2 (maintenance only)
Score: 27/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 76
- inputs: {"age_days": 1071, "days_push": 923, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1124, forks 79 (observed 2026-08-28T04:03:40.669470+00:00)

## What it is
ToRA is a series of tool-integrated reasoning LLM agents from Microsoft that solve mathematical reasoning problems by interleaving natural language reasoning with program-based tool use such as computation libraries and symbolic solvers. The repository provides research code, training pipelines, and released model checkpoints (7B to 70B) on Hugging Face.

## Use cases
- solve math word problems with llm agents
- train a language model to use tools for reasoning
- benchmark mathematical reasoning on MATH and GSM8k
- build tool-integrated reasoning agents
- fine-tune llms for program-aided problem solving
- reproduce ICLR 2024 tool-learning research

## When to choose
- you need state-of-the-art open-source models for competition-level math problem solving
- you are researching tool-integrated reasoning or tool learning for LLMs
- you want ready-to-use checkpoints fine-tuned for program-aided math reasoning

## When to avoid
- you need a production-ready math solver API rather than research code
- your task is general-purpose agent orchestration beyond mathematical reasoning
- you lack GPU resources for running or fine-tuning large models

## Facets
- artifact type: library
- maturity: maintenance
- function: agent-framework, llm-training, machine-learning, math
- domain: large-language-models, mathematics
- platform: python
- tags: mathematical-reasoning, tool-use, research-code, iclr-2024, huggingface-models, ai-agents, natural-language-processing, gpu

## Member repositories
- microsoft/ToRA (main) score 27

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:40.669470+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:39:50.475948+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/ToRA (fetched 2026-08-28T04:03:40.669470+00:00, sha cb193caf58d9)
  - homepage: https://microsoft.github.io/ToRA/ (fetched 2026-08-29T12:44:17.410751+00:00, sha b78d94beb3d6)
- Data as of 2026-08-30T08:39:29.467469+00:00.