GPT-6 Sol and Luna on Hermes Agent

Newer model: OpenAI released GPT-6.1 Sol on September 29. It costs the same per token as GPT-6 Sol with half-price cached input, and it also works through a ChatGPT subscription on Hermes Agent. Unlike GPT-6 Sol, it cannot turn reasoning off, so stay on GPT-6 Sol if your setup relies on reasoning none.

Two ways to run Sol and Luna

If you already pay for ChatGPT, the subscription route is the cheapest way to give an always-on agent GPT-6. OpenClaw Launch hosts Hermes Agent for you, so the agent keeps running on Telegram, Discord, WhatsApp, or web chat while your laptop is closed.

Deploy Hermes Agent See plans

Use Sol and Luna with your ChatGPT subscription

1. Deploy a Hermes Agent on OpenClaw Launch, or open an existing one from your Dashboard.
2. Click Terminal on the instance and run hermes auth add openai-codex --type oauth.
3. Open https://auth.openai.com/codex/device in your own browser, enter the code the terminal shows, and sign in with your ChatGPT account.
4. From any chat with your bot, switch the model:

   /model gpt-6-sol --provider openai-codex --global
   # or
   /model gpt-6-luna --provider openai-codex --global

Hermes reads the model list live from your ChatGPT account, so Sol and Luna show up in the Hermes model picker on the dashboard as soon as your plan serves them, without waiting for a release. The full walkthrough is in the Hermes ChatGPT subscription guide. Subscription usage counts against your ChatGPT plan's own limits.

What changed

OpenAI announced gpt-6-sol and gpt-6-luna on September 22. Both accept text and image input and return text. OpenAI describes Sol as the choice for complex coding and agent workflows, and Luna as its efficient choice for focused work at scale. They join GPT-6 Astra, which remains the highest capability tier.

Prices and limits

These are Sol and Luna Standard API rates for prompts up to 272,000 input tokens. Above that threshold, OpenAI applies 2× input and cache rates and 1.5× output rates to the whole request. Batch and Flex are half of Standard rates. A 1.05M-token window is a ceiling, not a reason to send every session that much context.

Both models support none, low, medium, high, xhigh, and max reasoning effort. OpenAI recommends the Responses API for tool use. In Chat Completions, Sol and Luna support function calling only when reasoning effort is none; use Responses when an agent needs reasoning and tools together.

Managed Hermes Agent with an API key

1. Add your OpenAI API key under API Keys.
2. Open your Hermes Agent model picker. If Sol or Luna is not listed, use the custom model fields: platform openai and model ID gpt-6-sol or gpt-6-luna.
3. Apply the change and start a new session. Your OpenAI API account must have access to the chosen model.

The managed Hermes custom model path uses the same model value for Sol and Luna.

Managed OpenClaw uses the same model picker and your OpenAI API key, but its runtime must also recognize the model. Choose Sol or Luna there only when it appears as an available OpenAI model for that bot; a custom ID entered before catalog support can fail on the first turn. The self-hosted registration steps below explain that difference.

Self-hosted Hermes Agent

Put the key in ~/.hermes/.env, then set provider and model separately in ~/.hermes/config.yaml. The example selects Sol; replace the bare ID with gpt-6-luna to use Luna.

# ~/.hermes/.env
OPENAI_API_KEY=sk-...

# ~/.hermes/config.yaml
model:
  provider: "openai-api"
  default: "gpt-6-sol"

Run hermes model if you prefer the provider wizard. Start a new session after changing the default. Hermes does not use the OpenClaw openai/ prefix in model.default. See the upstream configuration example.

Self-hosted OpenClaw

OpenClaw uses provider-prefixed references. Set up an OpenAI API-key auth profile, inspect your installed catalog, and then select a model that resolves:

openclaw onboard --auth-choice openai-api-key
openclaw models auth list --provider openai
openclaw models list --provider openai
openclaw models set openai/gpt-6-sol
# or: openclaw models set openai/gpt-6-luna

New model IDs can arrive before an OpenClaw release includes them in its local catalog. If the agent reports Unknown model, add the missing entries to your existing models.providers.openai.models array. The following openclaw.json excerpt selects Sol; change primary to openai/gpt-6-luna for Luna.

{
  "models": {
    "providers": {
      "openai": {
        "models": [
          { "id": "gpt-6-sol", "name": "GPT-6 Sol" },
          { "id": "gpt-6-luna", "name": "GPT-6 Luna" }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": { "primary": "openai/gpt-6-sol" },
      "models": {
        "openai/gpt-6-sol": { "agentRuntime": { "id": "openclaw" } },
        "openai/gpt-6-luna": { "agentRuntime": { "id": "openclaw" } }
      }
    }
  }
}

Merge these entries with any provider models and other settings you already have; do not replace the full config with the excerpt. The explicit agentRuntime.id keeps these examples on OpenClaw's embedded runtime. If you also have a Codex subscription profile, verify that the API-key profile is selected first in auth.order.openai for this agent. A successful config write does not prove API access. Check the upstream OpenAI setup and routing guide and model CLI guide for your installed version.

Keep API key billing separate from ChatGPT subscription access. For subscription-powered GPT-6 today, Hermes Agent is the direct route: it reads Sol and Luna from your ChatGPT account's live catalog. OpenClaw bakes its model catalog into each release, so check openclaw models list --provider openai-codex on your build before relying on a subscription model there.

Which model should you start with?

Choose Sol for a coding or research agent that has to plan and use tools across several steps. Try Luna for classification, extraction, short summaries, and frequent bounded automations. Test the same real tasks on both models and compare quality, latency, and total token use; per-token price alone does not measure cost per successful task.

For complex work where failures are costly, compare both with GPT-6 Astra. OpenAI's GPT-6 model guidance explains the family's tool and reasoning settings.

GPT-6 Sol and Luna FAQ

Related model guides

• GPT-6.1 Sol setup
• Hermes Agent with a ChatGPT subscription
• GPT-6 Astra setup
• Hermes Agent with your own API key
• OpenClaw with your own API key