# OpenBMB/ToolBench

[ICLR'24 spotlight] An open platform for training, serving, and evaluating large language model for tool learning.

Repository: https://github.com/OpenBMB/ToolBench
Canonical: https://ross.abutalabs.com/products/toolbench
Homepage: https://openbmb.github.io/ToolBench/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2025-05-21T15:46:59+00:00

## Health v2 (maintenance only)
Score: 39/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 22, release rhythm 35, longevity 85
- inputs: {"age_days": 1193, "days_push": 469, "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 5732, forks 489 (observed 2026-08-28T04:09:28.323624+00:00)

## What it is
ToolBench is an open-source platform for training, serving, and evaluating large language models on tool use, built around a large-scale instruction-tuning dataset covering thousands of real-world APIs. It includes the ToolLLaMA fine-tuned model, data generation pipelines, and the ToolEval automatic evaluator with a leaderboard.

## Use cases
- train an LLM to call real-world APIs
- evaluate tool-use capabilities of language models
- generate instruction tuning data for function calling
- benchmark my agent against ChatGPT on tool use
- fine-tune an open-source model for tool learning
- compare models on a tool-use leaderboard

## When to choose
- you need SFT data or a benchmark for LLM tool/API use
- you want to fine-tune or evaluate open models on function calling
- you need an automatic evaluator (pass rate / win rate) for tool-learning agents

## When to avoid
- you only need a production agent framework rather than training/evaluation research tooling
- you cannot access OpenAI APIs or the RapidAPI tool server
- you need a stable offline evaluation environment without setting up StableToolBench separately

## Facets
- artifact type: dataset
- maturity: active
- function: llm-training, agent-framework, benchmarking, rag
- domain: large-language-models, machine-learning
- platform: python
- tags: tool-learning, instruction-tuning, function-calling, api-use, tooleval, toolllama, iclr-2024, ai-agents, natural-language-processing, linux, docker

## Member repositories
- OpenBMB/ToolBench (main) score 39

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:28.323624+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 2026-08-29T17:53:23.346974+00:00, confidence not recorded.
  - readme: https://github.com/OpenBMB/ToolBench (fetched 2026-08-28T04:09:28.323624+00:00, sha 92cac552fef0)
  - homepage: https://openbmb.github.io/ToolBench/ (fetched 2026-08-29T08:48:48.782318+00:00, sha 8f626240f45f)
- Data as of 2026-08-30T08:39:29.467469+00:00.