GPT-6.2 Astra at $8 / $40, GPT-6.2 Sol at $2 / $10
Short answer. OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. Astra costs $8 input and $40 output per million tokens; Sol costs $2 and $10, the same figures as GPT-6.1 Sol. The same day GPT-6 Rosalind fell from $15 / $75 to $10 / $50, a promotional price “at least through December 31, 2026”.
In plain words. OpenAI has two new models. The bigger one costs a fifth less than GPT-6 Astra. The smaller one costs the same as the one before it. A token is a short piece of text, about three-quarters of a word. Prices are per million tokens.
Read on November 17, 2026, the day OpenAI published both models.
The two rows
OpenAI’s changelog entry of November 17 says: “Use GPT-6.2 Astra for demanding reasoning and coding tasks, or GPT-6.2 Sol for everyday coding and agent workflows. Both models support the Responses API and Chat Completions. Tool calling requires the Responses API.”
The prices come from the Markdown version of OpenAI’s pricing page, read the same day. Its Standard table is headed “Model | Short context input | Short context output | Short context cached input | Short context 30m write | Short context 12h write | Long context input | Long context output | Long context cached input | Long context 30m write | Long context 12h write”, and its first two rows are new:
- ”| gpt-6.2-astra | $8.00 | $40.00 | $0.40 | $10.00 | $20.00 | $16.00 | $60.00 | $0.80 | $20.00 | $40.00 |”
- ”| gpt-6.2-sol | $2.00 | $10.00 | $0.10 | $2.50 | $5.00 | $4.00 | $15.00 | $0.20 | $5.00 | $10.00 |”
Read by the header, that is:
| Per million tokens | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
| Input | $8.00 | $2.00 |
| Output | $40.00 | $10.00 |
| Cached input | $0.40 | $0.10 |
| Cache write, 30 minutes | $10.00 | $2.50 |
| Cache write, 12 hours | $20.00 | $5.00 |
| Input, prompts over 272K tokens | $16.00 | $4.00 |
| Output, prompts over 272K tokens | $60.00 | $15.00 |
Both model pages give “1,050,000 context window” and “128,000 max output tokens”, and the same rule for long prompts: “For requests with more than 272K input tokens, input and cache rates double and output costs 1.5 times the short-context rate for the full request.” The knowledge cutoffs differ: “May 18, 2026” for Astra, “Jun 01, 2026” for Sol.
GPT-6.2 Sol and GPT-6.1 Sol: the same ten figures
Two rows below GPT-6.2 Sol, the same table prints ”| gpt-6.1-sol | $2.00 | $10.00 | $0.10 | $2.50 | $5.00 | $4.00 | $15.00 | $0.20 | $5.00 | $10.00 |”. Every column matches, short context and long. For a buyer, then, the choice between the two Sol models is not about the bill. OpenAI’s pages do not say that GPT-6.2 Sol replaces GPT-6.1 Sol. Their deprecations page, read the same day, sets no shutdown for GPT-6.1 Sol and still names it as the recommended replacement for GPT-5 and o3, which shut down on December 11. We do not read more into the order of the rows.
GPT-6.2 Astra is different: it costs less than GPT-6 Astra, still listed at ”| gpt-6-astra | $10.00 | $50.00 | $1.00 | $12.50 | $25.00 |” for the short context. GPT-6 Astra costs 1.3× GPT-6.2 Astra on input and on output, and 2.5× on cached input ($1.00 against $0.40).
A month of requests
Take 1,000,000 requests a month, each with 2,000 input tokens and 500 output tokens, without caching: 2 billion input tokens and 500 million output tokens. The Claude and Gemini prices are from our dataset, checked on their providers’ pages on November 17, 2026: Anthropic’s table prints “Claude Fable 5.2 For demanding reasoning and long-horizon agentic work | $8 / MTok | $40 / MTok” and “Claude Sonnet 5.5 The best combination of speed and intelligence | $2 / MTok | $10 / MTok”, and Google’s prints “$1.50” input and “$7.50” output for Gemini 4 Hydrogen on its paid tier.
| Model | Input | Output | 1M requests, 2,000 in + 500 out |
|---|---|---|---|
| GPT-6 Astra | $10.00 | $50.00 | $45,000 |
| GPT-6.2 Astra | $8.00 | $40.00 | $36,000 |
| Claude Fable 5.2 | $8.00 | $40.00 | $36,000 |
| GPT-6.2 Sol | $2.00 | $10.00 | $9,000 |
| GPT-6.1 Sol | $2.00 | $10.00 | $9,000 |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $9,000 |
| Gemini 4 Hydrogen | $1.50 | $7.50 | $6,750 |
GPT-6 Astra’s month costs 1.3× GPT-6.2 Astra’s, $9,000 more. GPT-6.2 Astra’s costs 4.0× GPT-6.2 Sol’s. GPT-6.2 Sol’s costs 1.3× Gemini 4 Hydrogen’s.
Three pairs come out level on this bill: GPT-6.2 Astra with Claude Fable 5.2, and GPT-6.2 Sol with both GPT-6.1 Sol and Claude Sonnet 5.5. They part on cached input. Fable 5.2’s cache read is $0.20 against Astra’s $0.40, so with caching Fable 5.2 comes out cheaper. Sonnet 5.5’s is $0.10, the same as GPT-6.2 Sol’s. How the answers compare is a question for a test on real requests; the pages we read give no figure for it.
A long prompt changes the order
The level pairs split once a prompt passes 272,000 tokens, because of the models in this table only OpenAI’s charge more there. Anthropic’s pricing page says “Claude 4.6 and later models (except Claude Haiku 5.5) and Claude Mythos Preview include the full 1M token context window at standard pricing.” Google’s Gemini 4 Hydrogen page, read on November 13, says the model “uses flat standard pricing on the paid tier across the full 1M context window”.
Take 10,000 requests a month of 300,000 input tokens and 2,000 output tokens, without caching: 3 billion input tokens and 20 million output tokens.
| Model | Input | Output | 10,000 requests, 300,000 in + 2,000 out |
|---|---|---|---|
| GPT-6.2 Astra, prompts over 272K | $16.00 | $60.00 | $49,200 |
| Claude Fable 5.2 | $8.00 | $40.00 | $24,800 |
| GPT-6.2 Sol, prompts over 272K | $4.00 | $15.00 | $12,300 |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $6,200 |
| Gemini 4 Hydrogen | $1.50 | $7.50 | $4,650 |
Here GPT-6.2 Astra costs 2.0× Claude Fable 5.2, and GPT-6.2 Sol 2.0× Claude Sonnet 5.5 and 2.6× Gemini 4 Hydrogen.
Batch, Fast, Ultrafast and background windows
The model pages say: “Batch costs 50% of Standard rates. Fast mode costs twice the Standard rate for the same context length.” The pricing page’s Batch table prints $4.00 / $20.00 for Astra and $1.00 / $5.00 for Sol; its Fast table $16.00 / $80.00 and $4.00 / $20.00. Its Flex table has no row for either model.
Ultrafast is OpenAI’s fastest tier. Its guide, read on November 17, says “It supports GPT-6.2 Astra, GPT-6.2 Sol, GPT-6 Astra, and GPT-6.1 Sol.” The pricing page sets a launch price for the two new models: “For GPT-6.2 Astra and GPT-6.2 Sol, Ultrafast costs 3 times the Standard rate through February 28, 2027, and 6 times the Standard rate starting March 1, 2027 (Pacific time).” Its tables print $24.00 / $120.00 for Astra and $6.00 / $30.00 for Sol until then, and $48.00 / $240.00 and $12.00 / $60.00 after. GPT-6.1 Sol’s Ultrafast row is $12.00 / $60.00, so from March 1 GPT-6.2 Sol’s Ultrafast matches it, and until then costs half.
The second change of the day is a cheaper way to wait. The changelog says: “Choose a two-hour or eight-hour background completion window in the Responses API public beta. Model token rates are 30% lower than Standard for the two-hour window and 40% lower for the eight-hour window.” The pricing page names “GPT‑6.2 Astra, GPT‑6.2 Sol, GPT‑6.1 Terra, and GPT‑6 Luna”. It is a beta with conditions, such as “Discounted windows require Standard processing.”, not a list price, so our tables keep the Standard figure.
GPT-6 Rosalind: from $15 / $75 to $10 / $50
GPT-6 Rosalind is OpenAI’s restricted life-sciences model, listed under “Life Sciences — Trusted Access: Rosalind”. On November 16, 2026 the pricing page printed “gpt-6-rosalind | $15.00 | $75.00 | $1.50 | $18.75 | $37.50 |”. On November 17, in a table headed “Category | Model | Input | Output | Cached input | 30m write | 12h write”, it printed “gpt-6-rosalind (short context) | $10.00 | $50.00 | $1.00 | $12.50 | $25.00 |” and “gpt-6-rosalind (long context) | $20.00 | $75.00 | $2.00 | $25.00 | $50.00 |”. The old price is 1.5× the new one on input and on output. These are the same ten figures as GPT-6 Astra’s.
Under the table: “GPT-6 Astra for Law, GPT-6.2 Astra for Law, and GPT-6 Rosalind promotional pricing is available at least through December 31, 2026.” The words “at least through” set a floor, not a firm end date, and OpenAI publishes no price for after the promotion, so we show none. The same paragraph adds: “For standalone FedRAMP endpoints, GPT-6 Rosalind rates are 10% above the corresponding global rates.” OpenAI’s changelog does not date the cut, so our price-changes log records it on November 17, the first day we saw it.
How we checked
On November 17, 2026 we read OpenAI’s changelog, the Markdown version of its pricing page, the model pages of GPT-6.2 Astra and GPT-6.2 Sol and the Ultrafast and background guides, by script, and Anthropic’s and Google’s pricing pages for the comparison rows. Every quote above is from those readings. The prices of November 16 come from that day’s weekly check. The two new models are rows in our dataset from November 17; “gpt-6.2-astra-law”, which the pricing page lists as a legal “model configuration” at GPT-6.2 Astra’s price, is not a row of its own. The monthly bills are our own arithmetic from the per-million prices, in US dollars, at Standard, without caching, Batch or background windows; ratios are the larger bill divided by the smaller.
Frequently asked questions
How much do GPT-6.2 Astra and GPT-6.2 Sol cost?
OpenAI's pricing page, read on November 17, 2026, prints 'gpt-6.2-astra | $8.00 | $40.00 | $0.40 | $10.00 | $20.00' and 'gpt-6.2-sol | $2.00 | $10.00 | $0.10 | $2.50 | $5.00' per million tokens: input, output, cached input and the 30-minute and 12-hour cache writes. Above 272,000 input tokens the whole request costs $16.00 / $60.00 for Astra and $4.00 / $15.00 for Sol.
Does GPT-6.2 Sol cost the same as GPT-6.1 Sol?
Yes, on every column of OpenAI's Standard table read on November 17, 2026, short and long context, the two rows print the same ten figures, starting with $2.00 input and $10.00 output. OpenAI's pages do not say that GPT-6.2 Sol replaces GPT-6.1 Sol, and GPT-6.1 Sol is still listed.
What happened to GPT-6 Rosalind's price on November 17, 2026?
OpenAI's pricing page went from $15.00 input and $75.00 output per million tokens, read on November 16, to $10.00 and $50.00, read on November 17, and says 'GPT-6 Rosalind promotional pricing is available at least through December 31, 2026.' That is a floor, not a firm end date, and OpenAI publishes no price for after it.
Drafted with AI assistance from the cited source and reviewed by Álvaro Lucero Rufino, who is responsible for what stays published here. Automated publishing was discontinued in September 2026; posts that no longer met that standard were removed rather than left up. This is original explanatory writing, not a reproduction — read the source above for the full story.