GPT-6 Astra Pricing and Access: API Rates Plus ChatGPT Tiers

Coffee Summary

  • API model id `gpt-6-astra`: FACT rates $10 input / $50 output per 1M tokens; cached input $1; cache writes $12.50.
  • Context 1,050,000; max output 128,000; knowledge cutoff Apr 30, 2026. reasoning.effort: low / medium / high / xhigh / max.
  • Long-prompt surcharge FACT: prompts with >272K input tokens bill 2× input/cache and 1.5× output for the full request. Batch/Flex = 50% of Standard; Fast mode = 2×. Free API tier not supported.
  • ChatGPT access FACT (launch post): after the Sep 3, 2026 launch, Astra rolls out to Plus / Pro / Business / Enterprise over coming days; Enterprise admin enablement is off by default (work-focused OpenAI posts).
  • Secondary CLAIM (verify live Help Center): Plus may get Astra in Work/Codex before ordinary Chat; GPT-6 Pro chat allowances are tiered — do not promise plan limits from memory.

What happened

OpenAI launched GPT-6 Astra as its frontier model for hard end-to-end work (reasoning, coding, computer use, research, documents). This guide focuses on API pricing and who gets ChatGPT access — not a Critical Cyber builder checklist (see prior AI12-E02 coverage for that lane).

Primary API facts below come from OpenAI’s GPT-6 Astra model page and pricing notes. ChatGPT tier rollout mixes the launch post with Help Center / secondary reporting, labeled where uncertain.

Why it matters

Astra is expensive at list rates, and the >272K long-context multiplier can silently double input cost on the entire request. Teams that budget only “$10 / $50” without cache writes, Fast mode, or long-prompt rules will miss the real bill. On the product side, “available in ChatGPT” does not mean every seat sees the same surface (Chat vs Work/Codex) or the same weekly message bank.

What changed

API pricing FACTS (`gpt-6-astra`)

| Meter | Standard rate (per 1M tokens) |

| — | — |

| Input | $10.00 |

| Cached input | $1.00 |

| Cache writes | $12.50 |

| Output | $50.00 |

Additional FACTS from the model page:

  • Context window 1,050,000; max output 128,000; knowledge cutoff Apr 30, 2026.
  • reasoning.effort supports low, medium, high, xhigh, and max.
  • >272K input tokens: 2× input and cache rates and 1.5× output for the full request.
  • Batch and Flex: 50% of Standard. Fast mode: 2× applicable rates.
  • Free API tier: not supported for this model (Tier 1+ rate limits apply).

Modalities on the model card: text in/out, image input only; audio/video not supported on this card’s modality list.

ChatGPT access (launch + careful secondary)

| Audience | What sources say | Label |

| — | — | — |

| Plus / Pro / Business / Enterprise | Rolling out over days after Sep 3 launch | FACT (launch post) |

| Enterprise | Admin enable; **off by default** at launch | FACT (OpenAI work/access posts) |

| Plus vs Pro Chat | Plus may get Astra in **Work/Codex** while GPT-6 Pro Chat is framed for Pro/Business/Enterprise | CLAIM — verify Help Center |

| Message banks | GPT-6 Pro chat allowances are plan-tiered (e.g. Pro $100 vs $200 weekly banks in Help Center tables) | CLAIM — verify live limits before promising |

Do not treat secondary allowance tables as frozen contracts; OpenAI Help Center pages change. Confirm the live article before you tell finance “Plus includes N Astra chats.”

Who should care

  • Platform and finance owners modeling Astra API spend (cache, long context, Fast vs Batch).
  • Workspace admins deciding whether to enable Astra for Enterprise.
  • Plus/Pro buyers who need the difference between Work/Codex access and GPT-6 Pro Chat allowances.

Limitations

  • List prices exclude tool-call fees that may apply for search/computer-use style tools on other pages — check the live pricing page for tool meters.
  • ChatGPT allowance CLAIMs above must be re-verified; this article will not invent weekly caps.
  • Knowledge cutoff Apr 30, 2026 means post-cutoff world facts need tools/browsing, not parametric memory.
  • This pack intentionally skips Critical Cyber builder playbooks.

What to do next

1. Price a golden-path prompt at Standard rates, then re-price with cache hits, Batch/Flex, and a >272K input scenario.

2. Decide whether Fast mode’s 2× premium is worth latency for your UX; default many batch jobs to Batch/Flex.

3. Enterprise: leave Astra off until policy, DLP, and seat training are ready; then enable explicitly.

4. For ChatGPT seats, open the current Help Center pages for GPT-6 Pro / Work & Codex and screenshot the allowance table into your runbook.

5. Separate API budgets from ChatGPT seat budgets — they are not interchangeable.

AIImpish Take

Astra’s API card is clear and premium: $10/$50 with sharp long-context and Fast-mode multipliers that punish naive budgeting. Access is broader than “API only,” but ChatGPT rollout is seat- and surface-specific — Plus Work/Codex vs Pro Chat banks are easy to conflate. Cost the full request shape, enable Enterprise deliberately, and verify Help Center limits the day you promise them to users.