Topics / #benchmarks

Benchmarks projects on GitHub

As of , the most-starred Benchmarks projects on GitHub are dgl (14.3K stars), awesome-harness-engineering (4.4K stars) and ai-agents-the-definitive-guide (2.8K stars). GitHub.Locker tracks 8 Benchmarks repositories with 26.8K stars combined. The fastest-growing on October 19 was folkbench, with +100 stars in a day.

8Projects
26.8KStars in total
+123Stars in a day
Browse in the catalog →github.com/topics/benchmarks ↗

Hot right now

Most stars gained in a day · Oct 19

+100

folkbench

Folkbench/folkbench · ★ 2K

Folkbench is a benchmark and directory site for AI API relay providers (中转站) that resell access to models like GPT, Claude, Gemini, Grok and DeepSeek. It tracks each provider's availability rate, P95 response time, and pricing per model so you can compare them with real data instead of marketing claims.

+10

rrsi

google-research/rrsi · ★ 1.6K

RRSI (Regularized Recursive Self-Improvement) is a Google research framework that automatically improves an LLM agent's harness (prompts, control flow, tools, memory, context management around a frozen model) by searching over harness edits and keeping only those with measured, auditable gains.

+7

awesome-harness-engineering

walkinglabs/awesome-harness-engineering · ★ 4.4K

A curated list of articles, playbooks, benchmarks, specifications, and open-source projects about harness engineering — shaping the environment around AI agents so they can work reliably.

+4

ai-agents-the-definitive-guide

Nicolepcx/ai-agents-the-definitive-guide · ★ 2.8K

It is the companion code repository for the book 'AI Agents - The Definitive Guide': a set of runnable Jupyter notebooks that walk through every chapter, from basic agent patterns to production deployment and threat modeling.

The all-time top

8 projects · by stars

  1. #1 · dmlc/dgl

    dgl

    DGL is an open-source Python package for deep learning on graphs: it makes it easy to build and train graph neural networks on top of deep learning frameworks you may already use.

    14.3K
  2. #2 · walkinglabs/awesome-harness-engineering

    awesome-harness-engineering

    A curated list of articles, playbooks, benchmarks, specifications, and open-source projects about harness engineering — shaping the environment around AI agents so they can work reliably.

    4.4K
  3. #3 · Nicolepcx/ai-agents-the-definitive-guide

    ai-agents-the-definitive-guide

    It is the companion code repository for the book 'AI Agents - The Definitive Guide': a set of runnable Jupyter notebooks that walk through every chapter, from basic agent patterns to production deployment and threat modeling.

    2.8K
  1. 4

    folkbench

    2KFolkbench/folkbench · Rust

    Folkbench is a benchmark and directory site for AI API relay providers (中转站) that resell access to models like GPT, Claude, Gemini, Grok and DeepSeek. It tracks each provider's availability rate, P95 response time, and pricing per model so you can compare them with real data instead of marketing claims.

  2. 5

    rrsi

    1.6Kgoogle-research/rrsi · Python

    RRSI (Regularized Recursive Self-Improvement) is a Google research framework that automatically improves an LLM agent's harness (prompts, control flow, tools, memory, context management around a frozen model) by searching over harness edits and keeping only those with measured, auditable gains.

  3. 6

    cpu-performance-engineering

    1.2Kusamahz/cpu-performance-engineering · Python

    A curated reading path for CPU performance engineering: it walks you from what one instruction does on a modern core to serving ML inference on CPUs, ordered so each piece makes the next one legible.

  4. 7

    TurboBench

    374powturbo/TurboBench · C++

    TurboBench is a data compression benchmark: a single executable that tests 100+ compressors in memory and shows which ones win on speed and compression ratio.

  5. 8

    Hyperloom

    246AMD-AGI/Hyperloom · Python

    A multi-agent harness that autonomously optimizes LLM inference on AMD Instinct GPUs: it profiles your workload, searches framework and GPU-kernel optimizations, validates every candidate end to end, and saves proven results into a recipe knowledge base.

Often tagged together