# google/BIG-bench

Beyond the Imitation Game collaborative benchmark for measuring and extrapolating the capabilities of language models

Repository: https://github.com/google/BIG-bench
Canonical: https://ross.abutalabs.com/products/big-bench
Language: Python
License: Apache-2.0
License Family: permissive
Archived: true
Last push: 2024-07-19T11:57:37+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2056, "days_push": 775, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3247, forks 618 (observed 2026-08-28T04:07:50.691880+00:00)

## What it is
BIG-bench is a collaborative benchmark with over 200 tasks for probing and measuring the capabilities of large language models. It includes a Python evaluation harness, a lightweight 24-task subset (BIG-bench Lite), and a public leaderboard for comparing model performance.

## Use cases
- evaluate large language models on diverse tasks
- benchmark my LLM against other models
- find tasks that probe LLM reasoning capabilities
- compare language model performance on a leaderboard
- create custom benchmark tasks for language models
- measure how LLM capabilities scale with model size

## When to choose
- you need a broad, community-accepted suite of tasks to evaluate LLMs
- you want to compare your model against published results on a standard leaderboard
- you are researching how language model capabilities extrapolate with scale

## When to avoid
- you need a lightweight, fast evaluation of a single narrow capability
- you want a benchmark focused only on the newest frontier models or agentic tasks
- you need production inference tooling rather than evaluation

## Facets
- artifact type: dataset
- maturity: maintenance
- function: benchmarking, machine-learning, nlp, llm-inference, testing
- domain: large-language-models, machine-learning, artificial-intelligence, tutorials
- platform: python, cross-platform
- tags: llm-evaluation, benchmark, language-models, evaluation-suite, leaderboard, natural-language-processing

## Member repositories
- google/BIG-bench (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:50.691880+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 2026-08-30T07:24:28.179520+00:00, confidence not recorded.
  - readme: https://github.com/google/BIG-bench (fetched 2026-08-28T04:07:50.691880+00:00, sha e26b61757498)
- Data as of 2026-08-30T08:39:29.467469+00:00.