Is GPT-6 Astra new this week? Sume has listed it since September
The OpenAI page for GPT-6 Astra states no release date. Sume's registry carries Astra with a rate card dated 2026-09-05, so it is not a new pick for Sume users.
Short answer
I could not confirm a 2026-10-07 date for GPT-6 Astra. The OpenAI model page, read 2026-10-08, does not state a release date. What the Sume repo shows is older: the Astra row, tracked as #6053, has a rate card whose source is stamped openai_gpt_6_astra_list_2026-09-05, and the comment says OpenAI listed the standard rates as of that date. If something about Astra changed on October 7, it is not on the model page, and I make no claim about it.
For a Sume user the practical answer is that Astra is already a pick, enabled, never recommended and never the default.
What the page and the repo agree on
The prices on the model page today match the card in the repo, so there is no price change to react to.
| Item | OpenAI page | Sume card |
|---|---|---|
| Input per 1M | $10 | $10 |
| Cached input per 1M | $1 | $1 |
| Cache write per 1M | $12.50 | $12.50 |
| Output per 1M | $50 | $50 |
| Context window | 1,050,000 tokens | 1,050,000 (context constant in the repo) |
| Rate-card date | not stated | 2026-09-05 |
Why it matters
Launch-week articles recycle dates. If you are budgeting from a blog that says Astra launched this week, check the vendor page and your own receipts instead. In Sume, the Format run model field takes the catalog id gpt-6-astra, and the receipt shows the id that ran.
If a real change did land on October 7, such as a new tier or limit, the quickest check is the OpenAI pricing page; Sume's rate cards only change when someone updates the repo.
What it costs
On the standard card, a 30,000-input and 2,000-output turn is 30,000 x 10 / 1M = $0.30 plus 2,000 x 50 / 1M = $0.10, so $0.40. With the Fast checkbox, which Sume prices at 2x, it is $0.80.
What to verify instead
If you only have a launch-week post to go on, verify three items yourself. The price on the OpenAI model page: $10 input, $1 cached input, $12.50 cache write and $50 output per million as read on 2026-10-08. The effort options on the same page: low, medium, high, xhigh and max, with no default stated. And the context window, 1,050,000 tokens with 128,000 max output.
If any of those differ when you read the page, the Sume card is the stale side and the Format run receipt is the thing to trust for what you were charged.
There is a practical reason to care about the difference between a launch date and a listing date. Teams that read a launch post often assume a new model is the cheapest or fastest route to better output and rush a migration. In Sume's case, the Astra row has been selectable for weeks, so the question to ask is not whether it is available but whether the extra cost per turn buys you anything on your own Formats. That is a measurement you can make this week with two runs on identical inputs, using the A/B approach in the related post below.
- Price, effort options and context are all on one OpenAI page.
- The page lists web search, file search, MCP, computer use and tool search among supported tools.
Sources
Related posts
More in Models
- GPT Image 2.5 draft grid: four low-quality images for 4 cents
Four GPT Image 2.5 drafts in one Sume call at quality low cost 4 cents; a dollar buys 25 such calls. The arithmetic, the n field and when to go to high.
- Grok Imagine on Sume: 3 cents, one image per call, 9:20 phone ratios
Grok Imagine bills 3 cents per image on Sume, returns one image per call, lists 13 ratios including 9:20 and 19.5:9 phone shapes, and supports edits.
- H3 on vLLM-Omni: 87 s on 4 B300 for one clip, and the cost
MiniMax says an 8.7 s H3 clip takes about 87 s on 4 B300 GPUs. Turn that into GPU-seconds, a break-even rate, and compare with a 9 s Sume job.
- Haiku 4.5 in Sume's registry now points to Haiku 5.5: $1 to $0.10
Sume's registry retired Haiku 4.5 onto Haiku 5.5 with the successor set to always. List input drops from $1 to $0.10 per million; old rows keep the 4.5 card.
Written by Sume