Last verified
OPENAI FLAGSHIPGPT-6 FAMILY1.05M CONTEXTCHEAPER THAN GPT-6 ASTRA272K PRICE CLIFF

GPT-6.2 Astra API Pricing

GPT-6.2 Astra is OpenAI's flagship model, released November 17, 2026 at $8.00/M input, $0.40/M cached input and $40.00/M output. Every rate sits below the GPT-6 Astra it replaces at the top of OpenAI's list, which charges $10.00 and $50.00, and the cache read falls to 5% of input. GPT-6 Astra stays on the price list. Prompts over 272K input tokens bill on the long-context card at $16.00/M input and $60.00/M output.

Input - per 1M tokens
$8.00/M
Short context up to 272K input standard
Output - per 1M tokens
$40.00/M
Long context output is $60/M standard
Cached input - 95% off
$0.40/M
Cache writes cost $10.00/M -95%
Effective - agentic blend
$5.16/M
92/8 split - 82% cache
§ 01 / TERMINAL

GPT-6.2 Astra API cost calculator.

Live calculator pre-loaded with GPT-6.2 Astra's standard short-context rates. Reasoning effort runs from low to max - the API rejects none - and reasoning tokens bill as output, so put them in the reasoning field. Cross 272K input tokens in one prompt and the whole request moves to the long-context card.

$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
Open full calculator (all models · share URL · CSV) →
§ 02 / SCENARIOS

GPT-6.2 Astra cost on real workloads.

§ 03 / TOKENIZER

GPT-6.2 Astra token counter and cost.

What's a token →
Exact · o200k_base Auto-counts as you type

Counted with the o200k_base BPE encoding — the same tokenizer the API uses — entirely in your browser. The encoder loads on your first keystroke (~1 MB, one time); your text never leaves this page.

Characters 586
Words 97
Tokens (exact) 144 tokens
Cost as input · uncached $0.00115 USD
Cost as output · uncached $0.00576 USD
Cost as cached input $0.00006 USD
§ 04 / SHELF

GPT-6.2 Astra vs similar models.

All models →
Model Input /M Output /M Effective blended Context Best for
GPT-6.2 Astra Current $8.00 cache $0.40 $40.00 $5.16 agentic 92/8 1.05M OpenAI's flagship for the hardest work
GPT-6.2 Sol $2.00 cache $0.10 $10.00 $1.29 −4.0× 1.05M Everyday coding and agents at a quarter of the rate
GPT-6 Astra $10.00 cache $1.00 $50.00 $6.82 +32% 1.05M The previous Astra, still on the price list
GPT-6.1 Terra $1.00 cache $0.05 $5.00 $0.645 −8.0× 1.05M Visual reasoning, computer use and writing
GPT-6 Luna $0.10 cache $0.01 $0.50 $0.0682 −76× 1.05M High-volume routing tier
Claude Fable 5.2 $8.00 cache $0.20 $40.00 $5.01 −3% 1M Anthropic's top tier at the same list price
Gemini 4 Carbon $2.00 cache $0.10 $10.00 $1.21 −4.3× · promo 1M Google's top reasoning model
§ 05 / DEEP LINKS

GPT-6.2 Astra cost by scenario.

All calculators →
§ 07 / BENCHMARKS

GPT-6.2 Astra benchmarks

All benchmarks →

GPT-6.2 Astra has published scores on 3 benchmarks we track, and leads GPQA Diamond and shares first place on ExploitBench. All of them are the lab's own reports; each row links to the page that published it.

Benchmark Score Rank Leader Source
GPQA Diamond Expert knowledge 97% no tools, system card #1 of 13 GPT-6.2 Astra leads Vendor-reported OpenAI
ExploitBench Cybersecurity 100% V8, mean capability credit, five attempts #1 of 9 Tied for first with 2 more Vendor-reported OpenAI
SEC-bench Pro Cybersecurity 86.8% pass@1, maximum reasoning effort #4 of 11 Gemini 4 Neon · 90.5% Vendor-reported OpenAI

GPT-6.2 Astra API pricing FAQ.

GPT-6.2 Astra pricing questions - the standard card, what changed from GPT-6 Astra, the tiers it does and does not offer, and how it sits against GPT-6.2 Sol and Claude Fable 5.2.

Q · 01 How much does GPT-6.2 Astra cost? +
OpenAI lists gpt-6.2-astra at $8.00/M input, $0.40/M cached input, $10.00/M cache writes and $40.00/M output for prompts up to 272K input tokens. A 12-hour cache write costs twice the 30-minute write. Under our standard blend (92% input, 82% cache hits) that works out to $5.16/M.
Q · 02 Is it cheaper than GPT-6 Astra? +
Yes, on every line. GPT-6 Astra lists at $10.00 input, $1.00 cached and $50.00 output. 6.2 Astra charges $8.00, $0.40 and $40.00, so the cache-heavy agent blend falls from $6.82/M to $5.16/M. The cache read moves from a tenth of input to a twentieth. GPT-6 Astra stays on the price list and is on no deprecation schedule.
Q · 03 What happens above 272K input tokens? +
The whole request moves to the long-context card, not just the tokens past the line: $16.00/M input, $0.80/M cached input and $60.00/M output. Input and cache rates double and output rises by half.
Q · 04 Does GPT-6.2 Astra have Batch, Flex and Fast mode? +
Batch, yes, at half standard: $4.00/M input and $20.00/M output. Fast mode, yes, at double standard: $16.00/M and $80.00/M. Flex, no: OpenAI's Flex table lists GPT-6 Astra but not either GPT-6.2 model. In the Responses API, background completion windows in public beta bill 30% below standard for a two-hour window and 40% below for an eight-hour window.
Q · 05 What does Ultrafast mode cost? +
$24.00/M input and $120.00/M output today, three times standard. OpenAI says that rate runs through February 28, 2027 and that from March 1, 2027 Ultrafast on GPT-6.2 costs six times standard, the multiple GPT-6 Astra already pays at $60.00 / $300.00.
Q · 06 How does it compare with GPT-6.2 Sol? +
GPT-6.2 Sol lists at $2.00 / $10.00, a quarter of Astra on both sides, and blends to $1.29/M. OpenAI's split is Astra for demanding reasoning and coding and Sol for everyday coding and agent workflows.
Q · 07 How does it compare with Claude Fable 5.2? +
Claude Fable 5.2, Anthropic's top tier, has the same input and output sticker, $8.00 / $40.00. Its cache read is lower, $0.20/M against $0.40/M, so the cache-heavy blends come to $5.01/M and $5.16/M. Fable 5.2 has no long-context card above 272K input tokens.
Q · 08 What is GPT-6.2 Astra Pro? +
Some resellers list a "GPT-6.2 Astra Pro". OpenAI's price list, model pages and model list carry no such id; the model you can buy from OpenAI is gpt-6.2-astra.
Q · 09 What are the specs? +
1.05M context window, of which up to 922K can be input, with 128K max output and a knowledge cutoff of May 18, 2026. It takes text and image input and returns text, on the Responses and Chat Completions APIs; tool calling needs the Responses API. Regional data-residency endpoints cost 10% more.