OpenCompass

OpenCompass

Shanghai AI Lab's open-source LLM...

4.0· Editor Rating
China

Editor Review

8 capabilities, 29 tasks for comprehensive evaluation. CompassRank: LLM and MMBench multimodal leaderboards. Subjective + objective. Best for LLM developers, multimodal researchers.

AI Tools Navigator Editorial Team

What is OpenCompass

OpenCompass is Shanghai AI Lab's open-source LLM evaluation platform. CompassKit, CompassHub, CompassRank. 29 core tasks, 100+ datasets. Evaluates 8 capabilities: language, knowledge, logic, creativity, math, code, long text, agents.

Basic Info

Category:
Country:China

Best For

DevelopersResearchers

Difficulty: Intermediate

OpenCompass Key Features

  • Eight Capability Evaluation

    Language, knowledge, logic, creativity, math, code, long text, agents

  • 100+ Datasets

    29 tasks, 100+ datasets, subjective and objective

  • MMBench Multimodal

    Official multimodal leaderboard for VLMs

OpenCompass Key Advantages

  • 8 capabilities 29 tasks comprehensive
  • CompassRank + MMBench dual leaderboards
  • Open-source toolkit for reproducibility

OpenCompass Use Cases

  • Full LLM Evaluation

    Developers evaluate models across 8 dimensions

  • Multimodal Ranking

    Rank VLMs via MMBench and related benchmarks

OpenCompass Pricing & Credits

Pricing may change. Please refer to the official website for the latest pricing details.

Frequently Asked Questions

▶OpenCompass vs MMBench?

MMBench is OpenCompass multimodal subset for VLM evaluation.

▶How to evaluate with OpenCompass?

Use CompassKit, configure model and datasets for local or cloud run.

User Reviews

Real reviews and feedback from users