Open research · Model benchmarks

Know which models work best for coding

Independent benchmarks across intelligence, hallucination rates, and coding agent performance for coding model selection.

View Model Benchmarks Packages

103

Models tracked

September 16, 2026

Last updated

Benchmark dimensions

OpenRouter · AA

Data sources

Benchmarks

Leaderboard snapshot

Top models by Artificial Analysis Intelligence Index. Last updated September 16, 2026.

Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)

Anthropic

Intelligence Index

AA-Omni Accuracy: 67.2%low
Coding Agent Index: 81.6

Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback)

Anthropic

Intelligence Index

AA-Omni Accuracy: 66.2%low
Coding Agent Index: 80.7

GPT-6 Astra (max)

OpenAI

Intelligence Index

AA-Omni Accuracy: 62.6%low
Coding Agent Index: 76.9

View full benchmark table

What we build

Open, data-driven tooling for developers working with generative AI — starting with model selection benchmarks.

Model benchmarks

Living index of pricing, intelligence scores, hallucination rates, and coding agent performance.

Explore benchmarks

Composable packages

Coming soon

Early-stage @ck-ai/* monorepo packages — filesystem helpers, HTTP gateways, and SDK primitives for AI workflows.

Research & evaluation

Coming soon

Composable generative-AI tasks, experiments, and model evaluation pipelines — promote stable work into shared packages.

Packages

Early-stage libraries in the CK-AI monorepo — work-in-progress primitives for AI-powered developer tools.

In development

@ck-ai/sdk

Composable filesystem and agent-tool helpers for local workspaces.

listFilesTreereadTextFiles

In development

@ck-ai/gateway

HTTP gateway server — entry point for running services against the monorepo.

Bun runtime

In development

@ck-ai/research

R&D: composable generative-AI tasks, benchmarks, and model evaluation pipelines.

BenchmarksExperiments