Skip to content
StackLegder Software & AI-tool pricing — verified, dated, and logged when it changes
AI assistants & models
ANTHROPIC, OPENAI, GOOGLE

Claude Code vs Codex CLI vs Gemini CLI: the terminal coding agents, priced (2026)

Published Pricing verified

All three cost the same on a subscription — $20/month gets you the terminal agent — and all three can also run on an API key for pay-per-token billing. The choice is about which model ecosystem you want driving your terminal.

SubscriptionBacked byPower tiers
Claude CodeClaude Pro $20Opus 5 (default since late July 2026)Max $100 / $200
Codex CLIChatGPT Plus $20GPT-5.5 family, GPT-5.3-CodexPro $200
Gemini CLI / Antigravityfree tier, then $20GeminiUltra $100 / $200

Two things worth knowing. Claude Code ships with every paid Claude plan and now defaults to Opus 5 — and Claude still tends to lead independent coding benchmarks, so if raw code quality in the terminal is the goal, it's the front-runner. Codex CLI comes with ChatGPT Plus and also shows up in an IDE extension and SDK, so a $20 ChatGPT sub covers more surface than just the terminal.

Gemini's situation changed mid-2026: Google ended consumer authentication for the standalone Gemini CLI in June, folding it into "Antigravity" and routing paid access through Gemini Code Assist. There's still a free tier, then $20, then Ultra tiers that scale quota linearly.

If you already pay for one of the three assistants, use its terminal agent first — you're not saving money by adding a second. If you're choosing fresh: Claude Code for code quality and the Opus default, Codex CLI if you want the ChatGPT ecosystem and the IDE extension too, Gemini/Antigravity if you're in [Google Cloud](/entry/eks-vs-gke-vs-aks-managed-kubernetes-pricing-and-the-part-that-isnt-the-control-) and want the free tier to start.

Sources
  1. Anthropic — Claude Code
  2. OpenAI — Codex
  3. Google — Gemini CLI / Antigravity