GPT-6 Luna and Sol are now live on GreenTokens

GPT-6 Luna and GPT-6 Sol are available on GreenTokens from today. They complete the GPT-6 lineup alongside GPT-6 Astra, which has been available since 3 September, giving you three price tiers in the newest OpenAI generation: Luna at the low end, Sol in the middle and Astra at the top.

There's nothing to set up. If you already have a GreenTokens key, change the model in your request to gpt-6-luna or gpt-6-sol and you're using them.

Prices

Prices are per million tokens, next to OpenAI's official list price:

GPT-6 Sol is 60% below list across input, output and cached tokens. GPT-6 Luna is sold at its list price, which is already the lowest of any GPT model in our catalogue. Current prices for every model are always on the models page.

Model IDs

• gpt-6-luna: the lowest-priced GPT-6 tier.
• gpt-6-sol: the middle GPT-6 tier.
• gpt-6-astra: the top GPT-6 tier, available since 3 September.

All three accept images as input as well as text, so you can send screenshots, charts or photos in the same request as your prompt.

Try them now

GPT-6 models work through the OpenAI Chat Completions and Responses APIs, with streaming and tool calling, so the official OpenAI SDKs work unchanged:

from openai import OpenAI

client = OpenAI(base_url="https://api.greentokens.io/v1", api_key="sk-gt-…")

response = client.chat.completions.create(
    model="gpt-6-sol",
    messages=[{"role": "user", "content": "Summarise the key points of this release note."}],
)
print(response.choices[0].message.content)

The text API docs have the same call in TypeScript and curl, plus streaming, tools and image input. To use GPT-6 in Codex CLI, set model to gpt-6-sol in the provider config described on the Codex CLI page.

Compared with GPT-5.6

If you're on GPT-5.6 today, the new tiers are worth testing on price alone:

GPT-6 Luna costs less than half of GPT-5.6 Luna per output token, and GPT-6 Sol costs half of GPT-5.6 Sol. As with any model change, run your own prompts through the new model and compare results before switching production traffic; the ID is the only thing you need to change to test it.

Which GPT-6 tier to use

• Start high-volume, well-defined work, such as classification, extraction and short rewrites, on GPT-6 Luna.
• Use GPT-6 Sol where Luna's answers aren't good enough for your task.
• Keep GPT-6 Astra for the requests that need the top tier, and route everything else to the cheaper two.

Routing by difficulty like this keeps the average cost per request close to the lower tiers while still using Astra where it matters. Our guide to the cheapest OpenAI API model compares every GPT model's price in more detail.

Using GPT-6 in your tools

Because GPT-6 is served through the standard OpenAI formats, any tool that lets you set a custom OpenAI-compatible provider can use it. Point the tool at https://api.greentokens.io/v1 with your GreenTokens key, then use the model ID:

• Codex CLI: set model = "gpt-6-sol" in your GreenTokens provider config, or run codex --model gpt-6-luna for a single session.
• opencode: add gpt-6-sol and gpt-6-luna under your GreenTokens provider's models, then pick them with /models.
• OpenClaw: add them to your provider's models list and reference them as greentokens/gpt-6-sol.
• Hermes Agent: switch inside a session with /model custom:gpt-6-sol.

Switching safely

A new model can behave differently from the one you use today, even at the same tier. A simple way to switch without surprises:

1. Create a separate API key for testing, with a low daily spend limit.
2. Send a representative sample of your real prompts to the new model with that key.
3. Compare the answers with your current model, and check tokens and cost per request in the dashboard's Requests log.
4. Move traffic over gradually, starting with the requests where the new model did best.

Because every request is logged with its model, tokens and cost, you'll see the effect of the switch on your bill within a day, without waiting for an invoice.

Common questions

Do I need a new key or account?

No. Any existing GreenTokens key can call GPT-6 Luna and Sol straight away, billed from the same prepaid balance as every other model. If a key has spend limits, they apply to the new models too.

Is prompt caching supported?

Yes. Repeated prompt prefixes are cached automatically in the OpenAI format, and cached input is billed at a tenth of the normal input price: $0.01 per million tokens on Luna and $0.08 on Sol. Read more in what prompt caching is.

Get started

New to GreenTokens? Sign in with your email, top up a prepaid balance, create an API key and follow the quickstart. There's no subscription, and GPT-6 Luna, Sol and Astra sit alongside Claude, Grok, DeepSeek and the rest of the catalogue on one API.