Our take
Newless than six months old and changing fast- Activity
- 77
- Community
- 50
- Issues
- 83
- Pull requests
- 83
+111 stars in the last 30 days
- Merges more pull requests than 94% of the projects we track
Benchmark Radar is a broad, evidence-linked discovery resource with an active development pace, but its short history means teams should validate catalog coverage and score comparability for their use case.
Good fit if
- You need a central place to discover AI benchmarks and inspect linked evidence.
- Your research or evaluation work benefits from tracking reported scores and benchmark trends.
- You want either a web dashboard or a locally queried Python-based catalog.
Look elsewhere if
- You need a benchmark execution or scoring framework rather than a discovery catalog.
- Your workflow requires a guaranteed, independently verified source for every benchmark or score.
- You need permissive commercial reuse of the project's editorial content, which has separate restrictions from the MIT-licensed software.
All health signals
| Last commit | 2026-11-06 (today) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 14 (latest v0.14.0, 2026-11-03) |
| Contributors | 28 (top contributor: 75% of commits) |
| Issues closed, last 90 days | 35+ (typically closed in 1 day) |
| Pull requests merged, last 90 days | 91 (typically merged in 0 days) |
| Project age | 3 months |
Checked on 2026-11-07 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
406Forks
65Change since Nov 16, 2026: +2 stars·0 forks
Overview
Benchmark Radar is a Python-based, continuously updated catalog and search tool for AI benchmarks, evaluations, datasets, and data-quality records. It addresses the work of finding relevant evaluations, locating their datasets and code, and checking the conditions behind published scores by collecting records and evidence from public sources.
Researchers and evaluation engineers can use its web dashboard to discover recent findings, inspect benchmark trends and score histories, or query downloaded data with the CLI. The project also publishes its catalog data and offers an RSS feed, making it useful for both interactive discovery and research workflows.
Key Features
- Collects benchmark-related records and evidence from public sources, including research feeds, catalogs, and model reports.
- Provides a searchable benchmark catalog with source identities and citations.
- Tracks reported model scores, test conditions, and score histories where available.
- Shows recent discoveries in a ranked daily feed with a briefing that links to its evidence.
- Offers leaderboard, benchmark frontier, saturation, and trend views in the web dashboard.
- Publishes catalog and discovery data through Hugging Face and an RSS feed.
- Provides an offline CLI for querying downloaded data.
Use Cases
- Benchmark researchers can find relevant evaluations and follow new papers, datasets, and releases.
- Evaluation engineers can compare reported model scores and inspect their cited sources and test conditions.
- AI teams can monitor which benchmarks are attracting model evaluations and investigate possible saturation.
- Developers and research groups can query the catalog locally when they want benchmark discovery in a CLI workflow.
What you need
Detected in the repository
- Python >=3.11 (from pyproject.toml)
- A test suite and automated checks on GitHub Actions
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Open the dashboard for immediate use. To install the CLI skill for a coding agent, run:
npx skills add ktwu01/benchmark-radar
See the README and CLI guide for setup and usage details.
Considerations
The project is new and changing quickly, so its catalog and interfaces may evolve; verify records and citations before relying on them in consequential comparisons. Reported scores come from varied sources and conditions, and the project itself notes the limits of score comparisons. The software is MIT-licensed, while the technical report and original editorial content use CC BY-NC-SA 4.0 terms; commercial republication, resale, dataset packaging, and commercial product integration require prior written permission according to the README.
Found this useful?
Share it with someone who would like benchmark-radar.