Terminal / CLI

The #1 coding CLI agent. Leads SWE-bench Pro standardized (45.89%), wins independent head-to-heads on reasoning depth, ~4% of GitHub commits. Rate limits are the #1 complaint. Costs 2-3x more per task than Codex CLI due to higher token consumption.

Official · anthropics/claude-code

#5 coding CLI — active again after a development gap. v1.2.27 released 2026-03-16 with OpenAI official partnership (post-Anthropic block). 124,766 stars make this the largest AI coding repo by raw count. The unauthenticated RCE was fixed in v1.1.10+ but the corporate blocking incident and security history require explicit disclosure. Best for developers who want maximum model flexibility with a proven star-signal community.

Community · sst/opencode

#2 coding CLI. Terminal-Bench 77.3% (GPT-5.3-Codex) and 3-4x more token-efficient. Sandbox-first safety is a genuine differentiator. Free with ChatGPT removes the cost barrier. Trails Claude Code by ~5pp on SWE-bench Pro standardized (41.04% vs 45.89%).

Official · openai/codex

#3 coding CLI — best free entry point. 1K req/day free tier is unmatched, 1M context is the largest. SWE-bench Pro standardized 43.30% (#2, only 2.59pp behind Claude Code). File deletion incident (AI Incident Database #1178) and tool-calling weaknesses are real concerns.

Official · google-gemini/gemini-cli

#7 coding CLI — strong verification story (191K PyPI/week, 5.7M lifetime), but Codex CLI at 2.49M npm/week and Gemini CLI at 678K/week have overtaken Aider's download rank. Category pressure is real: HN thread 'Claude Code with Sonnet 4 is so good I have stopped using Aider' (#44154020). Best for Python devs who want fine-grained model control and git-native workflow. v0.86.2 (2026-02-12) is 5 weeks behind competitors shipping daily.

Community · Aider-AI/aider

Best pick for teams that want background AI agents enforcing code quality on PRs, not real-time autocomplete. The pivot repositioned it away from individual devs toward team CI workflows.

Community · continuedev/continue