Tooling
OpenAI GPT & ChatGPT Pricing in 2026: Every Model and Plan Explained
Every GPT API rate, the Codex coding agent, and the ChatGPT plans, with the cost levers that matter.
OpenAI’s 2026 pricing spans the new GPT-6 flagship, a wide GPT-5 lineup, the Codex coding agent, and ChatGPT subscription plans. This guide pulls the current numbers into one place.
Last checked: September 8, 2026, against OpenAI’s own pricing pages. Confirm before budgeting, since rates change often.
OpenAI API pricing by model
Rates are per million tokens, input/output. Cached input bills at 10% of the input rate across the lineup.
| Model | Input ($/M) | Output ($/M) | Best for |
|---|---|---|---|
| GPT-6 Astra | $10 | $50 | New flagship; hardest reasoning and agentic work |
| GPT-5.6 Sol | $4 | $20 | Production workhorse of the new generation |
| GPT-5.6 Terra | $2 | $12 | Budget mid-tier |
| GPT-5.6 Luna | $0.20 | $1.20 | Cheapest current model; volume tasks |
| GPT-5.5 | $5 | $30 | Prior flagship, still listed |
| GPT-5.5-pro | $30 | $180 | Maximum capability, premium price |
GPT-6 Astra now tops the lineup at $10/$50, exactly where GPT-5.5-pro used to sit for far more capability. If you were paying $30/$180 for 5.5-pro, moving to Astra is both an upgrade and a price cut.
The GPT-5.6 family (Sol, Terra, Luna) is the practical tier. Sol at $4/$20 handles most production workloads, and Luna at $0.20/$1.20 makes high-volume classification and extraction nearly free. Cached input at 10% of list rewards repeated-context workloads like chat apps and agents.
Older model pricing
Still listed and still answering plenty of search queries: GPT-5.4 at $2.50/$15, GPT-5.4 Mini at $0.75/$4.50, GPT-5.4 Nano at $0.20/$1.25, and GPT-5.3-Codex at $1.75/$14. Legacy budget models like GPT-4.1 nano (about $0.10/$0.40) remain available for simple, high-volume tasks.
Codex: now bundled into every plan
Codex is OpenAI’s cloud-based autonomous coding agent powered by the GPT-5 and GPT-6 families. It runs multi-step tasks in isolated sandboxes and works in parallel across projects, the direct competitor to Anthropic’s Claude Code.
The big change since June: the standalone Codex Plus ($20) and Codex Pro ($100) subscriptions no longer exist. Codex is now included in every ChatGPT plan, Free included, metered by credits. GPT-6 Astra consumes 250 credits per million input tokens; typical GPT-5.6 messages run 5 to 30 credits. For CI and automation, an API-key mode bills at standard API rates instead.
Cached input (repeated project context) still costs roughly 10% of the input rate, which remains the most important cost lever in any agentic workflow.
ChatGPT subscription plans
| Plan | Price | Notes |
|---|---|---|
| Free | $0 | Now includes Codex and ChatGPT Work access |
| ChatGPT Go | $8/mo | Budget tier, now available worldwide |
| ChatGPT Plus | $20/mo | Flagship access for individuals |
| ChatGPT Pro | from $100/mo | Two tiers: $100 for 5x limits, $200 for 20x |
| Business | $20/user/mo annual ($25 monthly) | Replaces the “Team” naming |
| Enterprise / Edu | Custom | Governance, security, scale |
Pro was restructured from a flat $200/month into two tiers, and “Team” became “Business” with a lower annual seat price.
Which OpenAI option should you pick?
For everyday assistant use, ChatGPT Plus is enough, and Go at $8 covers light use. Heavy Codex users should start at Pro $100 and only move to the $200 tier after hitting limits. On the API, default to GPT-5.6 Sol for production, drop to Terra or Luna for routine jobs, and reserve GPT-6 Astra for the work that genuinely needs it. Cache aggressively; it is the single biggest saving.
FAQ
How much does the OpenAI API cost in 2026? GPT-6 Astra $10/$50, GPT-5.6 Sol $4/$20, Terra $2/$12, Luna $0.20/$1.20, GPT-5.5 $5/$30 (pro $30/$180), GPT-5.4 $2.50/$15.
How much does ChatGPT cost per month? Go $8, Plus $20, Pro from $100 ($200 for 20x limits), Business $20/seat annual. Free now includes Codex.
What is the cheapest OpenAI model? GPT-5.6 Luna ($0.20/$1.20), with legacy GPT-4.1 nano cheaper for simple tasks.
How does Codex billing work? Bundled into every ChatGPT plan and metered in credits; the standalone Codex plans were discontinued. Cached input is ~10% of the input rate.
This page is part of our full AI model pricing and comparison guide. See also Claude pricing, Gemini pricing, and Grok pricing, or the best AI model for coding. Hunting for the floor? See the cheapest AI model.
Verify against the official OpenAI API pricing and ChatGPT plan pricing pages.
