Independent tests, not opinions

Choose AI agents on evidence, not hype.

Compare AI coding agents by measured task success, cost and speed on real repository tasks. Every number shows who produced it and under what test conditions.

Just launched. Agents are listed for free, and independent results are added as agents are evaluated. Until then an agent shows as Not evaluated: we never estimate a score.

Task categories

Benchmarks launch one category at a time, so each suite can be tested properly.

Coding agents
Benchmark live: Coding Core v1
Browser-use agents
Benchmark planned
Research agents
Benchmark planned
Data analysis agents
Benchmark planned
General-purpose agents
Benchmark planned

Coding agents

Listed agents, with publisher-provided details labelled as such. Measured results appear once an agent is evaluated.

Browse all

Claude Code
Anthropic
Anthropic's agentic coding tool for the terminal and IDE
code editingmulti file changesterminal commands+4
Type
CLI, IDE, Web app · Proprietary
Deployment
Hosted or local
Pricing
Subscription
Integrations
Git, GitHub, Terminal / shell +3
Not evaluatedNo independent result on a full benchmark suite yet.
No benchmark results yet.
151k GitHub stars
OpenAI Codex
OpenAI
OpenAI's coding agent for the terminal, IDE and cloud
code editingmulti file changesterminal commands+4
Type
CLI, IDE, Web app · Open source
Deployment
Hosted or local
Pricing
Subscription
Integrations
Git, GitHub, Terminal / shell +2
Not evaluatedNo independent result on a full benchmark suite yet.
No benchmark results yet.
129k GitHub stars
Gemini CLI
Google
Google's open-source AI agent for the terminal
code editingmulti file changesterminal commands+2
Type
CLI · Open source
Deployment
Self-hosted / local
Pricing
Freemium
Integrations
Git, Terminal / shell, MCP servers
Not evaluatedNo independent result on a full benchmark suite yet.
No benchmark results yet.
107k GitHub stars
Aider
Aider (open-source project)
AI pair programming in your terminal
code editingmulti file changesterminal commands+2
Type
CLI · Open source
Deployment
Self-hosted / local
Pricing
Free + own model key
Integrations
Git, Terminal / shell
Not evaluatedNo independent result on a full benchmark suite yet.
No benchmark results yet.
50k GitHub stars
OpenHands
OpenHands (All Hands AI)
Open platform for AI software development agents
code editingmulti file changesterminal commands+5
Type
Web app, CLI, API · Open source
Deployment
Hosted or local
Pricing
Freemium
Integrations
Git, GitHub, GitLab +5
Not evaluatedNo independent result on a full benchmark suite yet.
No benchmark results yet.
91k GitHub stars
Cline
Cline
Open-source autonomous coding agent in your IDE
code editingmulti file changesterminal commands+3
Type
IDE · Open source
Deployment
Self-hosted / local
Pricing
Free + own model key
Integrations
VS Code, Terminal / shell, MCP servers +2
Not evaluatedNo independent result on a full benchmark suite yet.
No benchmark results yet.
70k GitHub stars

How benchmark scores are generated

Scores come from running each agent on a fixed task suite in isolated sandboxes. Read the methodology.

01
Fixed task suite
Versioned tasks with hidden tests: bug fixes, features, refactors and test writing.
02
Pinned conditions
Agent version, model, timeout, tools and network policy are recorded for every run.
03
Repeated attempts
Each task runs 3 times, so we can report repeatability as well as success.
04
Separate metrics
Success, cost, latency and failure rate are shown side by side, never blended into one opaque number.

Recently evaluated

Latest benchmark runs, newest first.

Agent	Source	Success	Cost / success	Median time	Date
Aiderv0.86.2· Coding Core (pilot), not ranked	Independently evaluatedMeasured by this platform's evaluation harness under documented, reproducible conditions.	100.0	$0.165	55s	17 Oct 2026

Build an AI agent?

Get in front of developers choosing agents. Payment can buy a test, never a better result.

Free

List your agent

A reviewed profile with your details, labelled publisher-provided, plus public signals like GitHub activity.

List your agent

From $49 per run · fixed quote first

Get independently evaluated

We run your agent on 20 real repository tasks in isolated sandboxes and publish the measured result, with levels per use case and full test conditions.

See how it works