OpenAI · Checked October 13, 2026

How do I use GPT-6 Astra in Codex or OpenCode?

Q: How do I use GPT-6 Astra in Codex or OpenCode?

Standard Compute serves GPT-6 Astra as openai/gpt-6-astra on every monthly plan from $19, so Codex, OpenCode and Claude Code run it on a fixed compute budget that answers new requests with HTTP 402 at 100%. No OpenAI account or usage tier is involved.

Standard Compute serves GPT-6 Astra as openai/gpt-6-astra on every monthly plan from $19, so Codex, OpenCode and Claude Code run it on a fixed compute budget that answers new requests with HTTP 402 at 100%. No OpenAI account or usage tier is involved.

Model	OpenAI's top GPT-6 model, which OpenAI positions for its most demanding work.
Model ID	`openai/gpt-6-astra`Short form `gpt-6-astra` also works
Base URL	`https://api.stdcmpt.com/v1`Claude Code uses `https://api.stdcmpt.com`
Context	1.05M tokens at OpenAI, with 128K max outputA request here takes up to about 850K input and 50K output tokens
Input	Text and images
Reasoning	low, medium, high, xhigh or max, per request or as a Models page default

Copy a working config

Each config pins Astra for every request. Select Astra and GPT-6 Luna on the Models page and set the model to standardcompute to spend Astra only on the hard steps.

terminal/v1/responses

# 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-6-astra
standardcompute launch codex -- --model openai/gpt-6-astra

{
  "$schema": "https://opencode.ai/config.json",
  "model": "standardcompute/openai/gpt-6-astra",
  "provider": {
    "standardcompute": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Standard Compute",
      "options": {
        "baseURL": "https://api.stdcmpt.com/v1",
        "apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
      },
      "models": {
        "openai/gpt-6-astra": { "name": "GPT-6 Astra" },
        "standardcompute": { "name": "Smart routing" }
      }
    }
  }
}

# Install the Standard Compute launcher once
curl -fsSL https://standardcompute.com/cli/install.sh | sh

# Start Claude Code with openai/gpt-6-astra as the main model
standardcompute launch claude -- --model openai/gpt-6-astra

# Or configure it yourself. No /v1 here: Claude Code adds /v1/messages.
export ANTHROPIC_BASE_URL=https://api.stdcmpt.com
export ANTHROPIC_AUTH_TOKEN=YOUR_API_KEY
export ANTHROPIC_MODEL=openai/gpt-6-astra
# Set every slot: Claude Code's own claude-*-5-5 names return model_not_found
export ANTHROPIC_DEFAULT_OPUS_MODEL=openai/gpt-6-astra
export ANTHROPIC_DEFAULT_SONNET_MODEL=openai/gpt-6-astra
export ANTHROPIC_DEFAULT_HAIKU_MODEL=openai/gpt-6-astra
claude

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, Claude Code.

What it costs

Per 1M tokens	Prompt up to 272K	Prompt over 272K
Input	$10.00	$20.00
Cached input	$1.00	$2.00
Cache write	$12.50	$25.00
Output	$50.00	$75.00

OpenAI standard API rates; OpenAI counts a prompt over 272K input tokens as long context. Source: OpenAI API pricing, checked October 13, 2026.

A 90K-token Codex turn on GPT-6 Astra

11,700 fresh input × $10.00/M$0.1170

78,300 cache hits × $1.00/M$0.0783

700 output × $50.00/M$0.0350

One turn$0.2303

Average coding-agent turn on our gateway, Sep 13 to Oct 12, 2026: 87% of the prompt is cache hits and the reply is about 700 tokens. OpenAI list prices.

The $20 budget in the Starter plan covers roughly 90 turns like this one. Astra on every request for a working month needs a larger plan or a routed selection.

What works differently here

In Codex, built-in web search is off on this connection, and previous_response_id returns 400 because the Responses API here is stateless. Codex sends the full conversation each turn, so it works as shipped.
OpenAI's pro reasoning mode is not offered; the highest setting here is max.

Other ways to run GPT-6 Astra

OpenAI APIPay per token	First-party, with rate limits tied to your organization's usage tier.
ChatGPT plans with CodexSubscription	Codex's own sign-in, within each plan's usage limits, without a general API key for other agents.
OpenRouterPay per token	Served by OpenAI, with a batch variant at half price.
Standard ComputeMonthly plan	Astra next to GPT-6 Luna, the GPT-5.6 models, Claude and open models on one key.

FAQ

Do larger plans get a better version of GPT-6 Astra?

No. Plans differ in compute budget and request limits, not in models. Astra draws the budget down faster than mid-tier models because its list price is higher.

GPT-6 Astra or GPT-5.6 Sol?

GPT-5.6 Sol lists at $4 / $20 per million tokens, well under Astra. Select both and set the model to standardcompute, and routing decides per request which one answers.

Does GPT-6 Astra work in Claude Code?

Yes. The gateway serves every catalog model on the Messages API as well, so the Claude Code tab above works. Set Astra's reasoning level on the Models page; Claude Code's thinking toggle has no effect there.

Sources: OpenAI API pricing; OpenAI: GPT-6 Astra model page

Create your key

Get your API key