How do I use GPT-5.6 Luna in Codex or Hermes without per-token billing?
Put openai/gpt-5.6-luna in Codex, Hermes or OpenCode with a Standard Compute key: GPT-5.6 Luna comes with every monthly plan from $19, and usage draws on the plan's compute budget, which returns HTTP 402 at 100% with no per-token meter running. Its successor, openai/gpt-6-luna, is on the same key at half of GPT-5.6 Luna's list price.
| Model | The low-cost tier of OpenAI's GPT-5.6 family (July 9, 2026), aimed at cost-sensitive, high-volume work. It has a 1.05M-token context window and 128K max output. |
|---|---|
| Model ID | openai/gpt-5.6-lunaShort form gpt-5.6-luna also works |
| Base URL | https://api.stdcmpt.com/v1 |
| Input | Text, images and filesThe US pin openai/gpt-5.6-luna:us takes text only; the EU pin openai/gpt-5.6-luna:eu takes all three |
| Reasoning | minimal, low, medium or high |
| Plans | Every plan, from Starter at $19 a month |
Copy a working config
Luna's usual job is the many small calls around a stronger model, so routing is often the better setting: select Luna next to Terra or Sol on the Models page and use standardcompute as the model.
# Install Codex and the Standard Compute launcher once
npm install -g @openai/codex
curl -fsSL https://standardcompute.com/cli/install.sh | sh
# Start Codex on your plan with openai/gpt-5.6-luna
standardcompute launch codex -- --model openai/gpt-5.6-luna
Replace YOUR_API_KEY with a key from the dashboard. This config reads your key from STANDARDCOMPUTE_API_KEY; export it with a key from the dashboard. Full setup: Codex CLI, Hermes, OpenCode.
What it costs
| Per 1M tokens | GPT-5.6 Luna | GPT-6 Luna |
|---|---|---|
| Input | $0.20 | $0.10 |
| Cached input | $0.02 | $0.01 |
| Cache write | $0.25 | $0.125 |
| Output | $1.20 | $0.50 |
OpenAI standard API rates for prompts up to 272K tokens. Past that line GPT-5.6 Luna bills 2x input and 1.5x output on the whole request. Source: OpenAI API pricing, checked October 13, 2026.
At these rates, a month of Luna alone usually costs less on OpenAI's per-token API than any plan. People who pick a plan for Luna use it for the small calls around Terra, Sol or Claude on the same key.
What works differently here
- If you are choosing today, GPT-6 Luna lists at half the price and Artificial Analysis scores it one point higher (38 against 37). Keep GPT-5.6 Luna where an existing setup or eval depends on its behavior.
- OpenAI's xhigh, max and pro reasoning modes are not offered for Luna here.
Other ways to run GPT-5.6 Luna
| OpenAI API$0.20 / $1.20 per 1M | First-party, pay per token, with the long-prompt rate above 272K. |
|---|---|
| ChatGPT plans with CodexSubscription | Codex's native sign-in, within each plan's usage limits. |
| OpenRouter / AzurePay per token | Per-token access under another account, including OpenRouter's batch variant. |
| Standard ComputeMonthly plan | GPT-5.6 Luna and GPT-6 Luna side by side, with EU and US pins for 5.6. |
FAQ
Claude Haiku 4.5 or GPT-5.6 Luna?
On list price Luna costs a fifth as much: $0.20 / $1.20 per million tokens against $1 / $5 for Haiku 4.5. Haiku 4.5 is missing from this catalog because its 200K context is below the gateway's 1M minimum; Claude Haiku 5.5 takes its place.
GPT-5.3 Codex or GPT-5.6 Luna?
This gateway has no GPT-5.3 Codex. OpenAI lists it at $1.75 / $14 per million tokens, so Luna is far cheaper per token; for Codex sessions that need more than Luna, pin GPT-5.6 Terra or GPT-5.6 Sol.
Does GPT-5.6 Luna read images here?
Yes, on openai/gpt-5.6-luna and the EU pin openai/gpt-5.6-luna:eu. The US pin openai/gpt-5.6-luna:us is text-only.
Sources: OpenAI API pricing; OpenAI: GPT-5.6 Luna model page; Artificial Analysis: GPT-5.6 Luna
Plans from $19/mo
One key for GPT-5.6 Luna and the rest of the catalog, with a monthly compute budget you can watch in the dashboard and no overage charges.