Our take
Inactiveno commits in over a year- Activity
- 20
- Community
- 65
- Issues
- 21
- Pull requests
- 21
+0 stars in the last 8 days
This is a research-oriented toolkit with useful algorithmic abstractions and examples, but its inactive maintenance status means adopters should be prepared to maintain or verify it themselves.
Good fit if
- You are experimenting with search-based LLM reasoning and want reusable components rather than a single fixed prompting recipe.
- You need to inspect reasoning traces or adapt examples for planning, math, or browser tasks.
- You can work with Python 3.10 or later and are comfortable checking compatibility with your chosen model backend.
Look elsewhere if
- You need a project with recent commits, releases, or active issue and pull-request handling. The repository has had no commits in over a year and no release in the past year.
- Your team requires CI or a test suite as part of its adoption criteria. Neither is reported in the repository requirements data.
- You need a turnkey production agent rather than a library and examples for experimentation.
All health signals
| Last commit | 2025-06-10 (1.3 years ago) |
|---|---|
| Commits, last 90 days | 0 |
| Releases, last 12 months | 0 (latest v1.0.0, 2024-05-02) |
| Contributors | 22 (top contributor: 36% of commits) |
| Issues closed, last 90 days | 0 |
| Pull requests merged, last 90 days | 0 |
| Project age | 3 years |
Checked on 2026-10-08 with the GitHub API.
Overview
LLM Reasoners is a Python library for implementing and comparing methods that guide language models through multi-step reasoning. It separates a reasoning system into a world model, a search configuration with rewards, and a search algorithm, so developers can adapt these parts to a task rather than build the whole process from scratch.
It is aimed at researchers and engineers exploring approaches such as Monte Carlo Tree Search, Tree of Thoughts, and reward-model-guided inference. Its visualization tools can help inspect search traces, while support for several model backends lets teams experiment with different inference setups.
Key Features
- Reusable abstractions for world models, reward and search configuration, and search algorithms.
- Implementations and examples of methods including MCTS, BFS and DFS Tree of Thoughts, and PRM-guided inference-time scaling.
- Supports model backends including Hugging Face Transformers, OpenAI API, SGLang, Exllama, fairscale, and llama.cpp.
- Offers interactive visualization of reasoning traces and search trees.
- Includes examples across tasks such as web research, browser environments, math, and planning.
- Documents reproductions and analysis of selected published reasoning methods.
Use Cases
- Researchers can prototype search strategies for multi-step reasoning and compare them on their own benchmarks.
- LLM application developers can explore whether search or reward-guided inference improves a task beyond a single generation.
- Teams studying agent behavior can inspect search traces to understand how a reasoning process reached an answer.
- Developers building web or browser agents can use the included ReasonerAgent and BrowserGym examples as starting points.
What you need
Detected in the repository
- Python (from pyproject.toml)
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install the package with Python 3.10 or later:
pip install llm-reasoners
For repository examples and setup details, see the README.
Considerations
The project is marked inactive: it has no commits in over a year, no releases in the last year, and no recent recorded issue closures or merged pull requests. It has 22 contributors, with no measured concentration around one contributor, but that does not indicate current maintenance. The repository data also reports no CI or tests, so teams should validate their required workflows and backend compatibility before relying on it. The README notes that optional modules may need additional dependencies; the setup requirements can therefore vary by example and model backend.
Found this useful?
Share it with someone who would like llm-reasoners.