Is GPT-5.6 Sol available on a flat monthly API plan?
Yes: Standard Compute includes GPT-5.6 Sol (openai/gpt-5.6-sol) on every monthly plan from $19, through an OpenAI-compatible API with no OpenAI account or usage tier to unlock. Each plan is a fixed compute budget: at 100%, new requests get HTTP 402 and nothing is billed as overage.
| Model | The largest GPT-5.6 model, above Terra and Luna, with a 1.05M-token context window and 128K max output. |
|---|---|
| Model ID | openai/gpt-5.6-solShort form gpt-5.6-sol also works |
| Base URL | https://api.stdcmpt.com/v1Claude Code uses https://api.stdcmpt.com |
| Plans | Every plan, from Starter at $19 a month with a $20 compute budget |
| Request limits | Per plan and key, from 300 requests per minute on StarterYour OpenAI usage tier and spend history play no part |
| Per request | About 850K input and 50K output tokens |
| Reasoning | minimal, low, medium or high, per request or as a Models page default |
Copy a working config
Pick your agent. Each config pins Sol for every request. To use Sol only when a task needs it, select Sol and an efficient model such as GPT-6 Luna on the Models page and set the model to standardcompute: routing sends hard requests to Sol and routine ones to the efficient 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 pinned to GPT-5.6 Sol
standardcompute launch codex -- --model openai/gpt-5.6-sol
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, OpenCode, Hermes, Cline and other tools.
What it costs
| Per 1M tokens | Up to 272K | Over 272K |
|---|---|---|
| Input | $4.00 | $8.00 |
| Cached input | $0.40 | $0.80 |
| Cache write | $5.00 | $10.00 |
| Output | $20.00 | $30.00 |
OpenAI standard API rates by prompt size. OpenAI is running a promotional rate on Sol through at least November 21, 2026. Source: OpenAI: GPT-5.6 Sol model page, checked October 13, 2026.
Average coding-agent turn on our gateway, Sep 13 to Oct 12, 2026. OpenAI list prices.
Sol is one of the most expensive models in the catalog. The $20 Starter budget suits Sol on the hard steps; Sol on every request for a working month needs a larger plan.
What works differently here
- Sol runs on upstream provider capacity. When a provider turns a request away for load, the gateway retries it up to twice before your agent sees an error.
- Codex's built-in web search is off on this connection, and the launcher refuses
--config,--profileand--ossso it can keep the settings it writes. - The Responses API here is stateless:
previous_response_idreturns 400. Codex works as shipped because it sends the full conversation.
Other ways to run GPT-5.6 Sol
| OpenAI APIPay per token | First-party, metered from the first request. Rate limits follow your organization's usage tier, which rises with what it has spent. |
|---|---|
| ChatGPT plans with CodexSubscription | Sol in ChatGPT and Codex within each plan's usage limits, rather than a general API key for other agents. |
| Bedrock / Azure / OpenRouterPay per token | The same metered model under AWS, Azure or OpenRouter billing. |
| Standard ComputeMonthly plan | Sol, Terra, both Lunas and GPT-6 Astra on one key, next to Claude, Gemini and open models. |
FAQ
Does Sol access depend on my OpenAI usage tier or past spend?
No. You call Sol with a Standard Compute key, so a new account gets the same model on the same terms as an old one. The same models are on every plan; plans differ in budget and request limits.
Can I send only the hard steps to Sol?
Yes. Select GPT-5.6 Sol and an efficient model such as GPT-6 Luna on the Models page and set your agent's model to standardcompute. Routing sends hard requests to Sol and routine ones to the efficient model, so the budget lasts longer than with Sol pinned.
Can one key route between Fable 5 and GPT-5.6 Sol?
Yes, anthropic/claude-fable-5 and openai/gpt-5.6-sol are both in the catalog. Pin one with its model ID, or select both and let routing pick per request.
Does Sol work in Roo Code, Continue and Claude Code?
Yes. Roo Code and Continue take an OpenAI-compatible provider: base URL https://api.stdcmpt.com/v1, your key, and model openai/gpt-5.6-sol. Claude Code takes the same ID through the launcher's --model flag.
What happens when the budget runs out?
Requests already running finish, and new ones return budget_exhausted until renewal. Upgrading mid-month is prorated, and on Economy and above, optional Pay as you go adds compute at the usage rate plus 15%.
Are GPT-6 Sol or GPT-6.1 Sol available?
Not on Standard Compute yet; the newer OpenAI models served are GPT-6 Luna and GPT-6 Astra. On OpenAI's API, GPT-6 Sol and GPT-6.1 Sol list at $2 / $10 per million tokens.
Plans from $19/mo
One key for GPT-5.6 Sol and the rest of the catalog, with a monthly compute budget you can watch in the dashboard and no overage charges.