ASTRA PRICING
What does GPT-6 Astra cost?
The rate card is live. Your workload still determines the bill.
OpenAI now publishes model-level API rates for GPT-6 Astra. The rate card is translated into the units developers actually budget: fresh input, cached input, cache writes, output, processing mode, and large-context multipliers.
RELEASED · ACCESS EXPANDING
Use GPT-6 Astra through the route that fits.
Astra is available through multiple documented routes. ChatGPT, Work, Codex, API, Copilot, Foundry, and third-party catalog access each have their own plan, account, client, or provider rules.
A route listing is not a promise that your account, plan, workspace, or region is eligible.
Official OpenAI routes lead. GitHub and Microsoft routes are labeled separately.
HowToUseAstra does not sell access to Astra. Route listings are editorial and do not promise eligibility, identical product terms, or provider endorsement.
CURRENT ANSWER · September 20, 2026
$10 input · $50 output per million tokens.
Cached input is $1 per million tokens and cache writes are $12.50 per million. Access is route-specific, so a published rate card does not mean every account can use the model.
OFFICIAL API RATE CARD
Four token prices to keep separate.
Mixing cached and uncached input makes an estimate look precise while overstating the cost. Choose the right row for each token path in your workload.
| Billable unit | List price | Use it for |
|---|---|---|
| Input tokens | $10.00 / 1M | Standard input tokens |
| Cached input tokens | $1.00 / 1M | Prompt-cache hits |
| Cache-write tokens | $12.50 / 1M | Creating cached context |
| Output tokens | $50.00 / 1M | Generated output |
Source: OpenAI’s GPT-6 Astra model documentation ↗. Rates and access terms can change; re-check the official page before procurement or launch.
WHEN ANOTHER MODEL FITS BETTER
Choose by the job, not by the release headline.
Astra is the primary route. Use another model only when its workflow, access, or economics are a better fit.
GPT-5.6 Sol
The most direct stepping stone if your target is OpenAI’s tooling, model naming, and developer workflow.
- Why it fits
- OpenAI’s current model guidance identifies GPT-5.6 Sol as the full-capability GPT-5.6 route.
- Honest limit
- It is GPT-5.6 Sol, not GPT-6 Astra. Similar lineage does not establish Astra-level capability, access, or pricing.
Claude Fable 5.1
A strong candidate when the job is difficult coding, research, knowledge work, or a long-running agent task.
- Why it fits
- Anthropic describes Fable 5.1 as its most capable model for ambitious coding projects and knowledge work.
- Honest limit
- It is Anthropic’s model, not GPT-6 Astra. Safeguards, plan limits, retention, and actual task performance still need to be checked for your use case.
Fusion
A different answer to uncertainty: ask several models to analyze in parallel, then synthesize instead of relying on one model.
- Why it fits
- OpenRouter describes Fusion as a server-side tool for multi-model deliberation.
- Honest limit
- Fusion is an OpenRouter tool, not itself a frontier model and not GPT-6 Astra. Availability, model mix, data terms, and cost are separate.
These are editorial alternatives, not Astra and not paid placements. Their product terms and capabilities remain separate.
PRODUCT SOURCES
Product claims stay beside the action so the comparison remains inspectable.
PROVIDER ACCESS MAP
Where can you get Astra today?
A published rate card is not the same thing as an available route. This comparison keeps first-party, provider, and unverified paths separate so a budget decision does not smuggle in an access assumption.
| Provider | Access evidence | API / model field | Price signal | Best fit | Next step |
|---|---|---|---|---|---|
| OpenAI API first-party | Confirmed Confirmed in current first-party docs The model and API route are documented. Access remains account-specific through Trusted Access and eligible accounts. | gpt-6-astra Responses API · api.openai.com | $10 input · $1 cached · $12.50 cache writes · $50 output / 1M | The canonical model ID, first-party tools, and the cleanest migration path. | Check OpenAI access |
| OpenAI ChatGPT first-party | Confirmed Confirmed in current first-party docs OpenAI’s current Help Center names GPT-6 Pro, powered by Astra, for Pro $100, Pro $200, Business, and Enterprise; availability varies by product, plan, workspace, and rollout. | GPT-6 Pro / Astra ChatGPT model picker | Included within the applicable ChatGPT plan allowance; plan terms vary. | Interactive work without managing an API key. | Check ChatGPT access |
| OpenAI Work / Codex first-party | Confirmed Confirmed in current first-party docs OpenAI says Astra is rolling out in Work and Codex with plan and product-specific allowances; Codex clients must meet the current version requirements. | GPT-6 Pro / Astra ChatGPT Work · Codex desktop / CLI | Uses the applicable ChatGPT plan allowance; extra credits do not provide early access. | Long multi-step work and coding workflows in OpenAI’s products. | Check Work and Codex |
| GitHub Copilot first-party | Confirmed Confirmed in current first-party docs GitHub’s supported-model catalog lists GPT-6 Astra as GA; Copilot plan, client, and minimum-version requirements apply. | GPT-6 Astra Copilot model picker | Uses the applicable Copilot plan and premium-request terms. | Astra-assisted coding inside a supported GitHub Copilot client. | Check Copilot support |
| Microsoft Foundry cloud | Confirmed Confirmed in current first-party docs Microsoft’s current article says GPT-6 Astra is generally available in Microsoft Foundry; deployment, data zone, and account requirements still apply. | Provider display name Foundry Models · Foundry Agent Service | $10/$1/$12.50/$50 global short context; regional and long-context rates differ. | Enterprise identity, governance, data-zone choices, and Foundry-native workflows. | Read the Foundry route |
MONITORING LANES
OpenAI-compatible is not the same as Astra-supported.
These catalogs are useful places to recheck during the rollout, but the current review did not verify an exact GPT-6 Astra listing. Do not enter a key, pay a bill, or promise a customer access based on a generic compatibility claim.
Amazon Bedrock
UnknownNo exact GPT-6 Astra match was verified in the public model catalog checked on September 5, 2026.
Recheck AWS catalogOpenRouter
UnknownNo exact GPT-6 Astra listing was verified in the public model catalog checked on September 5, 2026.
Recheck model catalogTogether AI
UnknownNo exact GPT-6 Astra listing was verified in the public model library checked on September 5, 2026.
Recheck model libraryFireworks AI
UnknownNo exact GPT-6 Astra listing was verified in the public model library checked on September 5, 2026.
Recheck model libraryGroq
UnknownNo exact GPT-6 Astra match was verified in the public supported-model list checked on September 5, 2026.
Recheck supported modelsBaseten
UnknownNo exact GPT-6 Astra listing was verified on the public Model APIs page checked on September 5, 2026.
Recheck Model APIsHugging Face
UnknownNo exact GPT-6 Astra listing was verified in the public provider documentation checked on September 5, 2026.
Recheck provider docsRelationship disclosure: current provider and monitoring links are editorial and untagged. No paid placement or commercial ranking is implied.
PROCESSING AND CONTEXT
The headline rate is not the whole calculation.
The official model page documents processing multipliers and a large-context boundary. The calculator keeps them selectable so the assumption is visible in the result.
Standard
ConfirmedUse the list rates above.
The default comparison mode for ordinary requests.
Batch / Flex
Confirmed50% of the standard rate.
Use when the job fits the applicable throughput and latency terms.
Fast mode
Confirmed2× the standard rate.
The model guide says Fast mode has no latency SLA; EU data residency has additional compatibility limits.
Above 272K input
Confirmed2× input/cache · 1.5× output.
The multiplier applies to the full request under the current model-page rule.
ACCESS BEFORE BUDGET
A price is useful only when your account can reach it.
GPT-6 Astra is released through multiple routes. Keep access, spend, and workload assumptions as separate decisions.
- Confirm eligibility.Enterprise access is available through OpenAI’s Trusted Access Program for eligible organizations. Check the route-specific terms before budgeting.
- Choose the token mix.Estimate fresh input, cached input, cache writes, and output independently; reasoning effort can change the amount of output work.
- Choose the processing mode.Batch/Flex can reduce price when its terms fit. Fast mode costs more and is not a latency guarantee.
- Budget the whole workflow.Add tools, retrieval, image inputs, retries, concurrency, taxes, and human review outside the model-token estimate.
HISTORICAL CONTEXT
Do not confuse the $2,000 research figure with a rate.
Research compute and customer billing answer different questions. Keeping that line visible protects the calculator from turning an attention-grabbing number into a false quote.
TEN MATHEMATICAL RESULTS
Roughly $2,000 at GPT-5.6 Sol API rates.
OpenAI used this figure to describe the compute behind the research release. It is not GPT-6 Astra pricing, a subscription fee, or a forecast of your monthly spend.
PRICING FAQ
Budget the parts that move.
Answers use the current official model documentation checked September 20, 2026.
How much does GPT-6 Astra cost?
The current OpenAI model page lists $10 per million input tokens, $1 per million cached input tokens, $12.50 per million cache-write tokens, and $50 per million output tokens. Your bill depends on token mix, processing mode, tools, retries, and account terms.
Is Astra API pricing available to everyone?
Pricing is documented, but access is route-specific. OpenAI describes Trusted Access and eligible-account rollout rules; free API access is not supported on the current model page.
Does Astra charge separately for reasoning?
The current GPT-6 Astra model page lists input, cached input, cache writes, and output rates; it does not list a separate reasoning-token rate. Keep reasoning effort and token usage visible, and re-check the official rate card before committing a budget.
What changes above 272K input tokens?
For requests above the documented 272K input-token threshold, the model page lists 2× input and cached-input rates and 1.5× output rates for the full request. The calculator flags this as a boundary rather than silently applying a blended rate.
What is the $2,000 Astra number people mention?
OpenAI’s mathematics announcement describes roughly $2,000 of compute at GPT-5.6 Sol API rates for finding the ten published results. That is research-compute context, not an Astra API price or a forecast of a customer invoice.
SOURCE LEDGER
A rate card you can audit.
The current pricing and processing rules come from OpenAI’s model documentation. The research-compute comparison remains linked to the original mathematics announcement.
OWNED OFFER · FIXED-SCOPE REVIEW
Get an Astra Production Readiness Review.
Founding test: $299. For one planned workload, the review includes intake, one 60-minute working session, a written decision memo within two business days, and one 20-minute follow-up. Submit the workload first; if it is a fit, scope, timing, and payment are confirmed before either side commits.
No charge today. Do not submit secrets, source code, customer data, or regulated data. This is independent implementation guidance, not OpenAI sales or endorsement.