AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What GPT-6.2 Astra And Sol Cost And Mean For Work on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on Nov. 17, 2026, positioning Astra for demanding projects and lower-cost Sol for everyday work and agents. Artificial Analysis scores Astra just behind Claude Fable 5.2 and reports lower costs on its benchmark task, but coding-agent results did not improve and OpenAI rates both models Critical for cybersecurity.

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on Nov. 17, 2026, alongside a legal configuration called Astra for Law 6.2. Independent testing placed Astra near the top of the Artificial Analysis Intelligence Index and found lower costs per benchmark task than several rivals, while OpenAI rated both new models Critical for cybersecurity and the evaluator reported no improvement in its coding-agent index.

OpenAI describes Astra as its model for the most demanding projects and Sol as its option for everyday work, coding and agents at scale. Its developer guide recommends Astra for the hardest tasks and Sol for complex work at lower cost. Both models accept text and images, have a 1,050,000-token context window, and can return up to 128,000 output tokens. Fine-tuning is not supported.

Standard API prices are $8 per million input tokens and $40 per million output tokens for Astra, and $2 and $10, respectively, for Sol. Astra’s input and output rates are 20% below those of GPT-6 Astra, its predecessor. Artificial Analysis estimates that its index task costs $4.00 with Astra at Max effort and $1.01 with Sol at Max. The same task costs $26.54 with Claude Fable 5.2 at Max and $5.48 with Gemini 4 Carbon at High. These are costs for the evaluator’s task, not a forecast of every company’s bill.

On the Artificial Analysis Intelligence Index, Astra scored 62 at Max effort and Sol scored 57. Claude Fable 5.2 led with 65 in the rounded headline figures. More precise scores show Astra at 62.14 and Fable at 62.22, placing Astra third among individual configurations on its model page. Artificial Analysis said Astra improved five points over GPT-6 Astra and Sol improved two; it also reported that Sol’s score was on par with GPT-6 Astra at about 15% of the cost per task. The evaluator found no improvement in its Coding Agent Index and reported hallucination rates of 59% for Astra and 55% for Sol at Max effort.

At a glance
announcementWhen: Announced November 17, 2026; independen…
The developmentOpenAI launched GPT-6.2 Astra, GPT-6.2 Sol and a legal configuration, with independent tests showing competitive benchmark results, lower costs on one measured task and significant safety caveats.

GPT-6.2 Astra, Sol and Astra for Law

Three models, three jobs

What launched, what it costs and how it scores in independent tests. Astra and Sol at their Max setting; the better value in each row is outlined.

AstraSolFor Law Price in/out$8/$40$2/$10$8/$40 Index score6257– Cost per task$4.00$1.01– Minutes per task10.47.9– Legal test15.0%11.1%20.8% Both rated Critical for cyber risk
  1. Three releases, one dayOn November 17, OpenAI released GPT-6.2 Astra for the hardest projects and GPT-6.2 Sol for everyday work. Astra for Law 6.2 is a legal configuration at Astra’s price.
  2. The independent indexAstra scores 62, second among model families behind Claude Fable 5.2 at 65. Sol scores 57. One index task costs $4.00 and $1.01.
  3. Speed and a legal testAstra needs 10.4 minutes per task, Sol 7.9. On the Harvey LAB-AA legal test, Astra for Law passes 20.8% of tasks.
  4. OpenAI’s own safety ratingBoth models meet OpenAI’s Critical threshold for cybersecurity. Higher-risk cyber help now needs extra vetting.

Source: Artificial Analysis, model pages for GPT-6.2 Astra, GPT-6.2 Sol and Astra for Law 6.2, accessed Nov 17, 2026 (Intelligence Index, cost and time per task, Harvey LAB-AA v1.1); OpenAI, API pricing page and GPT-6.2 System Card, Nov 17, 2026. Prices per million input/output tokens. Redrawn by Thorsten Meyer AI.

Graphic: Powered by Thorsten Meyer AI

What the Prices Mean for Work

The launch gives employers and developers a new choice between higher capability and lower unit cost. On Artificial Analysis’s fixed task, Sol’s cost was $1.01, compared with $4 for Astra, while Astra scored five points higher on the evaluator’s index. That price-performance split could matter to organizations running many model calls, including agent workflows, where per-task costs can accumulate. It does not establish that Sol will be cheaper or more effective for every workload: actual spending depends on prompt length, output volume, speed tier and the work being done.

Published token rates also have conditions that can materially affect budgets. Requests exceeding 272,000 input tokens incur doubled input and cache rates and output at 1.5 times the short-context rate for the whole request. Batch processing costs half the Standard rate, while Fast costs twice as much. Ultrafast is listed at three times Standard through Feb. 28, 2027, then six times from March 1. Eligible background requests receive discounts of 30% for a two-hour completion window and 40% for an eight-hour window. Buyers need to compare expected usage and service requirements, not just headline rates.

For workplace adoption, the evaluation results are mixed rather than a blanket endorsement. A high overall index score does not guarantee better coding-agent performance or fewer unsupported answers. OpenAI’s cyber-risk rating also makes access controls and internal review relevant considerations for organizations handling sensitive work. The results support evaluating the models against a company’s own tasks and risk rules before moving important workflows onto them.

Products Worth Considering

Bose QuietComfort Ultra Wireless Noise Cancelling Earbuds, 2nd Gen, with 20W 2-Port Wall Charger

Bose QuietComfort Ultra Wireless Noise Cancelling Earbuds, 2nd Gen, with 20W 2-Port Wall Charger

As an affiliate, we earn on qualifying purchases.

How OpenAI Positions the Models

The release divides the lineup by intended use: Astra for demanding projects and Sol for complex work where lower cost matters. OpenAI lists Astra for Law 6.2 as a legal model configuration, with the same standard token prices as Astra. The available documentation cited in the source does not explain how the legal configuration differs from the general model. Artificial Analysis had not assigned it an Intelligence Index score, though it had tested it on one legal benchmark.

The models’ knowledge cutoffs differ: May 18, 2026, for Astra and June 1, 2026, for Sol. OpenAI’s developer documentation also flags an API behavior change: when conversation history is edited mid-conversation, GPT-6.2 drops its earlier reasoning. The company warns that answer quality can decline and the model may use more tokens repeating earlier work. Requests that set reasoning effort to “none” are rejected.

OpenAI lists rate limits that vary by account usage tier, from 5,000 requests and 1 million tokens per minute to 15,000 requests and 40 million tokens per minute on the tiers described. Approved organizations can use Zero Data Retention and US or EU data residency with Standard, Fast and Batch processing; Ultrafast is listed with US data residency only. The source report said OpenAI’s launch post was inaccessible, so it could not establish which ChatGPT plans include the models.

Open Questions on Risk and Access

OpenAI’s system card rates Astra and Sol at Critical for cybersecurity, the higher of the two risk thresholds used in the report. The source material says more permissive cyber access requires extra vetting, but does not detail the full access process or how it will be applied across customer types. OpenAI also says Astra is the first model it has released externally to meet its High threshold for AI self-improvement. These are the company’s risk classifications; the supplied material does not provide enough detail to independently assess the underlying evaluations.

OpenAI’s same-day safety report lists regressions as well as improvements, but the source material does not enumerate every finding or explain their size. It also does not establish how the benchmark scores translate into performance on specific jobs, or whether the reported hallucination rates apply beyond the evaluator’s stated test conditions. The source report did not have access to OpenAI’s launch post, leaving ChatGPT plan availability unconfirmed there. Pricing and rate limits may also differ with account tier, workload, and processing option.

What Buyers Should Check Next

Organizations considering the models can compare Astra and Sol on their own representative tasks, tracking quality, token use, latency and total cost rather than relying on one benchmark. They should also check the applicable token rates for long prompts and speed tiers, and review their requirements for cyber-related use and data residency. OpenAI’s documentation says promotional pricing for Astra for Law is available at least through Dec. 31, 2026; the cited material gives no later end date.

Further details may clarify how Astra for Law differs from the general model, which ChatGPT plans include the release, and how OpenAI applies additional cyber-access vetting. Artificial Analysis has not yet provided an Intelligence Index score for Astra for Law. Until those points are documented, the confirmed picture is a high-scoring model family with lower measured task costs than some rivals, alongside uneven benchmark results and elevated cyber-risk classifications.

Key Questions

How much do GPT-6.2 Astra and Sol cost?

Standard API rates are $8 per million input tokens and $40 per million output tokens for Astra. Sol costs $2 per million input tokens and $10 per million output tokens. Long prompts and faster processing tiers can raise the rates; eligible Batch and background requests can lower them.

Which model is better for everyday work?

OpenAI positions Sol for everyday work and lower-cost complex tasks, and Astra for the most demanding projects. Independent index scores favor Astra, but Artificial Analysis reported Sol at about 15% of Astra’s cost per task in its test. Results will vary by workload.

Did GPT-6.2 improve coding-agent performance?

Not in the measure reported here. Artificial Analysis said there was no improvement in its Coding Agent Index. That finding is limited to the evaluator’s index and does not establish how the models perform on every coding task.

What does the Critical cybersecurity rating mean?

OpenAI’s system card assigns both models its Critical cybersecurity rating, the higher of the two risk thresholds used in the report. The source material says more permissive cyber access requires extra vetting, but does not provide the complete process or all underlying evaluation details.

What is Astra for Law 6.2?

It is a legal model configuration listed by OpenAI at Astra’s standard API rates. The available source material does not explain how it differs from Astra, and Artificial Analysis had not assigned it an Intelligence Index score as of Nov. 17, 2026.

Source: ThorstenMeyerAI.com

You May Also Like

China’s DeepSeek V4 Pro Boosts AI Power — Is It Worth The Cost Compared To Claude Fable?

China’s DeepSeek claims its V4 Pro model is upgraded, with assertions that it’s only 5% less effective than Claude Fable, but at a fraction of the cost. Details remain unverified.

The Biggest Misconceptions About the AI Market Right Now

No, AI market dominance isn’t inevitable—discover how regulations and competition are shaping a different future worth exploring.

Why AI Security Is Now Central to Enterprise Strategy

Meta Description: Many enterprises now prioritize AI security to stay ahead of evolving threats, but the true impact on your organization’s resilience is yet to be revealed.

Signal: Three Gates Close In Nineteen Days — The Pre-Release Regime Goes Global

China, the US and EU are activating AI controls within 19 calendar dates, but their rules cover different systems and risks.