LLM Evaluation Resources

A curated list of frameworks, benchmarks, and tools for evaluating large language models and AI agents.

Type Badges

• Library — Python/JS library or SDK — use programmatically

• CLI — Command-line tool

• Server — Self-hosted web application with UI

• Desktop — Desktop GUI application

Evaluation Frameworks

General-purpose frameworks for building and running LLM evaluations.

• deepeval — Library · The LLM Evaluation Framework

• openevals — Library · Ready-made evaluators for your LLM apps

• evalite — CLI Desktop · Evaluate your LLM-powered apps with TypeScript (local dev UI)

• lighteval — Library CLI · All-in-one toolkit for evaluating LLMs

• evalscope — Library CLI · Framework for efficient LLM/VLM evaluation

• giskard-oss — Library Server · Open-source evaluation and testing library for LLM agents

• Eval — Library · High-performance LLM evaluation with parallel API calls

• simple-llm-eval — Library · Simple LLM evaluation using LLM-as-a-judge

• evidently — Library Server · Evaluate, test, and monitor ML and LLM-powered systems

• ai-eval — Library · Prompt evaluation and optimization system for LLM applications

• neuro-judge — Library · LLM-as-a-Judge framework with multi-model, multi-criteria, and cost tracking

• OpenJudge — Library · Unified framework for holistic evaluation and quality rewards

• One-Eval — Library · Automated LLM evaluation system via agents

• GAGE — Library CLI · Unified evaluation engine for LLMs, MLLMs, audio, and diffusion models

• lm-evaluation-harness — Library CLI · A framework for few-shot evaluation of language models

Benchmarks

Standardised benchmark suites and leaderboards for comparing model performance.

• opencompass — Library CLI · LLM evaluation platform supporting 100+ datasets

• VLMEvalKit — Library CLI · Evaluation toolkit for large multi-modality models

• olmes — Library CLI · Reproducible, flexible LLM evaluations

• bench — CLI · A tool for evaluating LLMs

• genai-bench — CLI · Token-level performance evaluation of LLM serving systems

• sparse-frontier — Library · Evaluation framework for training-free sparse attention in LLMs

Observability & Monitoring Platforms

End-to-end platforms combining evaluation with tracing, logging, and monitoring.

• langfuse — Server · Open-source LLM engineering platform with observability

• opik — Server · Debug, evaluate, and monitor your LLM applications

• trulens — Library Server · Evaluation and tracking for LLM experiments and AI agents

• langwatch — Server · Platform for LLM evaluations and AI agent testing

• openlit — Server · OpenTelemetry-native LLM observability, monitoring, and evaluations

• mlflow — Library Server · Open-source platform for agents, LLMs, and ML with debugging and monitoring

• agenta — Server · Open-source LLMOps platform: playground, management, evaluation, and observability

Security & Adversarial Testing

Tools focused on red-teaming, safety, and robustness testing.

• rogue — Library · Stress-test your AI agents before attackers do

• moonshot — Library Server · Evaluating and red-teaming LLM applications with benchmarking

• agent-security-sandbox — Library · Benchmark for indirect prompt injection defenses in LLM agents

• AgentDefense-Bench — Library · Security benchmark for infrastructure-layer defenses in MCP-based agent systems

Agentic Performance

Benchmarks and tools for evaluating LLMs acting as autonomous agents.

• AgentBench — Library CLI · Comprehensive benchmark to evaluate LLMs as agents

• claw-eval — Library · Human-verified evaluation harness for evaluating LLMs as agents

• strands-agents/evals — Library · Comprehensive evaluation framework for AI agents and LLM applications

• agentdojo — Library · Dynamic environment to evaluate attacks and defenses for LLM agents

• AgentCPM — Library · End-to-end infrastructure for training and evaluating LLM agents

• MemoryAgentBench — Library · Benchmark for evaluating memory in LLM agents via multi-turn interactions

• HaluMem — Library · Operation-level hallucination evaluation benchmark for agent memory systems

• HarnessLab — Library · Benchmark for evaluating LLM agent harness components (context, retry, memory)

• ResearchHarness — CLI · Trusted-local harness for research agents with real tool use and evaluation

• iris-eval/mcp-server — Server · Agent eval standard for MCP — quality scoring, safety, and cost budgets

Tool-Calling Evaluation

Tools for evaluating function calling and tool-use capabilities.

• gorilla — Library CLI · Training and evaluating LLMs for function calls

RAG Evaluation

Tools specifically designed for evaluating retrieval-augmented generation pipelines.

• ragas — Library · Supercharge your LLM application evaluations (RAG-focused)

• AutoRAG — Library Server · RAG AutoML tool for finding the optimal RAG pipeline

Domain-Specific Evaluation

Evaluation tools targeting specific domains or verticals.

• med-lm-envs — Library · Automated LLM evaluation suite for medical tasks

• MedEvalKit — Library · A unified medical evaluation framework

Utilities & Experimentation

Tools for model comparison, grid search, and inference optimization.

• ollama-grid-search — Desktop · Desktop app for evaluating and comparing multiple LLM models via grid search

• speculators — Library · Library for speculative decoding algorithms for LLM inference

Other Resource Lists

• llm-benchmark — Curated list of LLM evaluation frameworks and benchmarks