LiveBench
Contamination-limited benchmark with regularly released question sets. Model answers and judgments are on Hugging Face.
- Operator
- LiveBench team / Abacus AI
- Leaderboard
- livebench.ai
- Category
- General / multi-benchmark
- How results are made
- Operator runs the evals
- Checked
- 2026-10-31
Raw data
| Format | Where |
|---|---|
| HF dataset | huggingface.co/datasets/livebench/model_judgment Model judgments. “Also available are the [model answers](https://huggingface.co/datasets/livebench/model_answer) and the [model judgments]” (raw.githubusercontent.com) |
| HF dataset | huggingface.co/datasets/livebench/model_answer Model answers. |
Licence / reuse: Unclear
Apache-2.0 (copied from FastChat) (licence of the repository that holds the result files). A licence exists, but it covers the code or website rather than clearly the results.
“The original LICENSE from https://github.com/lm-sys/FastChat is copied below.”
Source: raw.githubusercontent.com/LiveBench/LiveBench/main/LICENSE
The repo LICENSE is the Apache 2.0 text copied from FastChat; GitHub reports NOASSERTION. The HF result datasets carry no licence field.
Submissions: Not stated
We found no statement about outside submissions.
Last updated: 2026-10-25 Updated in last 90 days
- 2026-10-25: Date stated on the page (“Last updated 25 October 2026”) · livebench.ai
This page lists where the data lives and what the licence says. It does not copy any scores; open the leaderboard for those. Not legal advice. Something wrong? Use the bug link in the footer. Back to all leaderboards.