# openai/frontier-evals

OpenAI Frontier Evals

Repository: https://github.com/openai/frontier-evals
Canonical: https://ross.abutalabs.com/products/frontier-evals
Language: Python
License: MIT
License Family: permissive
Last push: 2026-04-21T20:53:31+00:00

## Health v2 (maintenance only)
Score: 55/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 78, release rhythm 35, longevity 37
- inputs: {"age_days": 523, "days_push": 134, "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 1287, forks 177 (observed 2026-08-28T04:04:15.048672+00:00)

## What it is
A collection of benchmark evaluations from OpenAI for measuring frontier LLM capabilities, including PaperBench (AI paper replication), SWE-Lancer (freelance software engineering tasks), and EVMbench (smart contract security). Each eval is an isolated Python project managed with uv.

## Use cases
- benchmark frontier LLM capabilities
- evaluate models on real freelance software engineering tasks
- test whether models can replicate state-of-the-art AI papers
- assess model performance on smart contract security tasks
- reproduce published OpenAI eval results
- compare model performance across capability benchmarks

## When to choose
- you need standardized benchmarks for evaluating frontier LLMs
- you want to reproduce results from OpenAI's published eval papers
- you're researching model capabilities in coding, research replication, or security

## When to avoid
- you need a general-purpose eval framework for building custom benchmarks
- you want lightweight model testing rather than heavyweight research-grade evals
- your use case is outside the included eval domains

## Facets
- artifact type: dataset
- maturity: active
- function: benchmarking, testing
- domain: artificial-intelligence, large-language-models, developer-tools, security
- platform: python, cli
- tags: llm-evals, frontier-models, paperbench, swelancer, evmbench, model-capabilities, ai-safety, evaluation

## Member repositories
- openai/frontier-evals (main) score 55

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:15.048672+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 2026-08-30T04:56:06.336094+00:00, confidence not recorded.
  - readme: https://github.com/openai/frontier-evals (fetched 2026-08-28T04:04:15.048672+00:00, sha 5fce136159ab)
- Data as of 2026-08-30T08:39:29.467469+00:00.