11 tools · cloud & local rankings

Best Models for AI Coding Tools

Cloud-locked agents like Claude Code, Codex, and Cursor each drive a specific vendor's models — we rank them by capability, token efficiency, and cost. Self-hosted agents run local models, where tool-calling reliability decides the tier. Pick the right model for the tool you actually use.

Cloud-locked agents

Claude Code

Cloud

CLI agent

Anthropic's terminal coding agent — pointed at a local model

STop pick: Claude Opus 4.8→

Codex

Cloud

CLI agent

OpenAI's coding agent — ranked across the GPT-5 model lineup

STop pick: GPT-5.5→

[]

Cursor

Multi-vendor

AI code editor

The AI code editor — a curated cross-vendor frontier lineup

STop pick: Claude Opus 4.8→

Self-hosted (local LLM) agents

OpenCode

CLI agent

Open-source terminal coding agent (Ollama / OpenAI-compatible)

STop pick: Qwen3.5 27B Instruct→

🐾

OpenClaw

Self-hosted agent

Self-hosted agentic assistant with first-class Ollama support

STop pick: Qwen3.6 35B-A3B→

Aider

Terminal pair-programmer

Terminal AI pair-programmer with its own edit-format leaderboard

STop pick: Qwen2.5 Coder 14B→

{}

Cline

IDE extension

Autonomous coding agent for VS Code (formerly Claude Dev)

STop pick: Qwen3 30B→

Roo Code

IDE extension

Autonomous VS Code agent with modes (a Cline fork)

STop pick: Qwen3.6 35B-A3B→

Continue.dev

IDE extension

Open-source chat, edit & autocomplete for VS Code / JetBrains

STop pick: Qwen2.5 Coder 14B→

Goose

CLI agent

Block's open-source autonomous developer agent (MCP-driven)

STop pick: Qwen2.5 Coder 14B→

Open Claude Code

CLI agent

Open-source Claude Code CLI reimplementation, run on local models

STop pick: Qwen3.6 35B-A3B→

New open-weight models, real Apple Silicon benchmarks, and the one model worth running on your Mac this week. Free, one email a week, unsubscribe anytime.