Side by side · November 2026

GPT-6 Astra vs GPT-6.2 Astra

GPT-6 Astra and GPT-6.2 Astra are within three points on the Models at Work leaderboard, which we treat as a tie (59 vs 58 out of 100). On independent tests alone they are also level (69 vs 72). GPT-6.2 Astra is cheaper: 1,000 support replies cost about $40 at list price, against $50 on GPT-6 Astra, which is 1.3 times as much. On the Artificial Analysis Intelligence Index, one of the independent tests, GPT-6.2 Astra leads 62 to 57.

Updated Nov 18, 2026 · every number links to who measured it
GPT-6 Astra · OpenAIGPT-6.2 Astra · OpenAI
Leaderboard rank#27#28
Score (0–100)5958
Kind of workHardest problemsHardest problems
Price per 1M tokens$10 to read · $50 to write$8 to read · $40 to write
Reads at once1M tokens (about 750,000 words)1M tokens (about 750,000 words)
Where you can get itOpenAI API, Amazon Bedrock, Microsoft Foundry, Gemini Enterprise Agent Platform (Google Cloud), ChatGPT Pro, Business and EnterpriseOpenAI API, Amazon Bedrock, ChatGPT Plus, Pro, Enterprise and Edu
Can a company host it itselfNoNo
Artificial Analysis Intelligence Index5762
Humanity's Last Exam, Diamond (Scale)60.60Not measured
Humanity's Sixth Sense (Scale)53.60Not measured
FrontierMath Erdős (Epoch)3% (2 of 68)Not measured
Vals Finance Index55.64% (rank 30)57.31% (rank 22)
Vals Tax Agent Bench20.72% (rank 56)21.84% (rank 55)
Vals Legal Research Bench39.42% (rank 40)42.79% (rank 31)
Remote Labor Index (Scale)20.83% of projects automatedNot measured
Professional Reasoning Benchmark - Finance (Scale)47.54Not measured
Professional Reasoning Benchmark - Legal (Scale)48.36Not measured
Best forProblems where a wrong answer costs more than the bill, Research-grade analysis, Hard scientific and mathematical questionsProblems where a wrong answer costs more than the bill, Research-grade analysis, Reading long contracts and filings

Cost of 1,000 jobs

1,000 of theseGPT-6 AstraGPT-6.2 Astra
Support repliesthe customer message, the account notes and two help articles read; a 300-word reply written$50$40
Documents processeda two-page invoice or form read; the fields pulled out$35$28
Calls summariseda 40-minute call transcript read; a summary and next steps written$130$104
Contracts revieweda 40-page contract read against a checklist; a memo of issues written$375$300
Drafts writtena brief and one example read; a 600-word email or proposal section written$55$44

Worked out from each vendor’s list price and a fixed size for each job. A product built on a model charges its own price on top. See what every model costs for the same jobs.

Choose GPT-6 Astra for

  • Problems where a wrong answer costs more than the bill
  • Research-grade analysis
  • Hard scientific and mathematical questions

Superseded on 17 Nov by GPT-6.2 Astra, which lists 20% lower ($8 and $40 against $10 and $50) and scores 62 against 57 on Artificial Analysis's index. Still sold; OpenAI gives no retirement date.

Caveats: The price doubles on any request that feeds it more than 272,000 tokens of text at once, such as a stack of long documents. The slowest of the top-tier models in Artificial Analysis's speed measurement. In ChatGPT it is sold as GPT-6 Pro, on the Pro, Business and Enterprise plans; this is from our researcher's reading of OpenAI's help page, which blocked automated reading this run. Now the most expensive model here and no longer OpenAI's most capable; a team paying by usage should ask its vendor when it moves to GPT-6.2 Astra.

Choose GPT-6.2 Astra for

  • Problems where a wrong answer costs more than the bill
  • Research-grade analysis
  • Reading long contracts and filings

OpenAI's new top model: 62 on Artificial Analysis's index, second to Claude Fable 5.2's 65 at the same $8 and $40 list price, but on Vals AI's finance test it gets 57.3% right against 56.7% for GPT-6.2 Sol at a quarter of the price.

Caveats: The price rises on any request that feeds it more than 272,000 tokens of text at once: reading costs double and writing 1.5 times as much for the whole request, OpenAI's model page says. On Vals AI's tax test it gets 21.8% of tasks right, below GPT-6.2 Sol's 24.8%; on legal research 42.8% against Sol's 37.0%. The ChatGPT plans come from 9to5Mac's report of 17 Nov, which does not mention the Business plan; OpenAI's launch page refused automated reading. Microsoft Foundry, Google Cloud and Microsoft 365 Copilot are not listed because no page read on 18 Nov names GPT-6.2 there. Scores within a point of GPT-6 Astra, which the method treats as a tie: no user reports are scored yet, so its place rests on tests, price and where it is sold.

What people running them say

GPT-6 Astra: The launch drew 2026's largest thread about a model on Hacker News, a developer forum, and the stories that followed were about what it can do rather than what it costs: an Enigma-coded message unsolved since 2005, a First World War cipher, a driving test.

GPT-6.2 Astra: The launch thread on Hacker News, a developer forum, drew 1,283 points and 798 comments in a day, and the early replies are mixed: one commenter calls Sol 'a marginal step forward' after a day, another reports Astra refusing a task on hardware they own, others praise the speed against Gemini 4 Carbon. No reports yet from operations teams.

Questions people ask

Which is better, GPT-6 Astra or GPT-6.2 Astra?

GPT-6 Astra and GPT-6.2 Astra are within three points on the Models at Work leaderboard, which we treat as a tie (59 vs 58 out of 100). On independent tests alone they are also level (69 vs 72). GPT-6.2 Astra is cheaper: 1,000 support replies cost about $40 at list price, against $50 on GPT-6 Astra, which is 1.3 times as much. On the Artificial Analysis Intelligence Index, one of the independent tests, GPT-6.2 Astra leads 62 to 57.

When should I choose GPT-6 Astra over GPT-6.2 Astra?

Choose GPT-6 Astra for problems where a wrong answer costs more than the bill, research-grade analysis, hard scientific and mathematical questions. Superseded on 17 Nov by GPT-6.2 Astra, which lists 20% lower ($8 and $40 against $10 and $50) and scores 62 against 57 on Artificial Analysis's index. Still sold; OpenAI gives no retirement date.

When should I choose GPT-6.2 Astra over GPT-6 Astra?

Choose GPT-6.2 Astra for problems where a wrong answer costs more than the bill, research-grade analysis, reading long contracts and filings. OpenAI's new top model: 62 on Artificial Analysis's index, second to Claude Fable 5.2's 65 at the same $8 and $40 list price, but on Vals AI's finance test it gets 57.3% right against 56.7% for GPT-6.2 Sol at a quarter of the price.

Which is cheaper, GPT-6 Astra or GPT-6.2 Astra?

GPT-6 Astra costs $10 per million tokens it reads (input) and $50 per million tokens it writes (output) on OpenAI's list price (OpenAI API pricing, read Nov 18, 2026). A token is about three-quarters of a word. GPT-6.2 Astra costs $8 per million tokens it reads (input) and $40 per million tokens it writes (output) on OpenAI's list price (OpenAI API pricing, read Nov 18, 2026). A token is about three-quarters of a word.

What do 1,000 jobs cost on GPT-6 Astra and GPT-6.2 Astra?

At list price: 1,000 support replies cost $50 on GPT-6 Astra and $40 on GPT-6.2 Astra; 1,000 documents processed cost $35 on GPT-6 Astra and $28 on GPT-6.2 Astra; 1,000 calls summarised cost $130 on GPT-6 Astra and $104 on GPT-6.2 Astra; 1,000 contracts reviewed cost $375 on GPT-6 Astra and $300 on GPT-6.2 Astra; 1,000 drafts written cost $55 on GPT-6 Astra and $44 on GPT-6.2 Astra. A product built on either model charges its own price on top.

Scores come from the Models at Work leaderboard: Score is 0 to 100 and computed, not typed. 40% is independent testing (Artificial Analysis, Arena, Scale, Epoch, Vals and others; each test is scaled against the best score on this board, and tests of real work, such as finished freelance projects or professional tasks, count double the exams). 30% is price: the vendor's list price for a typical mix of reading and writing, on a fixed scale from $0.05 to $60 per million tokens, so a model that costs a fifth as much gains real ground. 15% is what people report after running it, from Signals and linked threads. 15% is fit: how much text it can take in at once and where a business can get it (the products and clouds that carry it), with a small credit for open weights. A missing component drops out and the rest are reweighted. A model with no independent test yet is provisional and ranks below every measured one. Within three points is a tie. Vendor-published figures are labelled and never counted. Prices are vendor list prices where published; a typical mix assumes three parts reading to one part writing.