🔍 Read the full analysis: GPT-6.2 Astra And Sol: What OpenAI’s New Models Cost, Score And Change For Work on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on Nov. 17, 2026, alongside a legal configuration of Astra. Artificial Analysis scored Astra 62 and Sol 57; its reported cost for one index task was $4.00 for Astra Max and $1.01 for Sol Max. OpenAI rates both models Critical for cybersecurity, while coding-agent results showed no improvement.
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on Nov. 17, 2026, positioning Astra for demanding projects and Sol for lower-cost everyday work. Independent tests published that day put Astra near the top of Artificial Analysis’s model-family index and found lower cost per benchmark task than several named rivals, while OpenAI’s safety report rated both models Critical for cybersecurity.
Artificial Analysis’s Intelligence Index gave Astra a score of 62 at its Max setting and Sol 57. Claude Fable 5.2 led the model-family comparison at 65. The margin is narrow when configurations are compared more precisely: Astra Max scored 62.14 and Fable 5.2 Xhigh scored 62.22. Astra’s model page ranked it third among 91 configurations, rather than second. These are results from the index, not OpenAI’s own benchmark claims.
Artificial Analysis reported a cost of $4.00 per index task for Astra Max and $1.01 for Sol Max. Its comparison listed $26.54 for Claude Fable 5.2 Max and $5.48 for Gemini 4 Carbon High. These task-cost estimates are distinct from API token prices and reflect the index’s workload and method; they should not be treated as a guaranteed price for a customer’s particular use.
OpenAI lists standard API rates of $8 per million input tokens and $40 per million output tokens for Astra, down from $10 and $50 for GPT-6 Astra. Sol costs $2 per million input tokens and $10 per million output tokens, unchanged from the previous Sol rates given in the source material. Cached input costs $0.40 per million tokens for Astra and $0.10 for Sol. Astra for Law 6.2 is also listed at Astra’s standard rates, with promotional pricing stated to run at least through Dec. 31, 2026.
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.
- 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.
- 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.
- 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.
- 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
Lower Task Costs, Serious Cyber Ratings
The launch gives developers a choice between a higher-capability model and a less expensive option, with the index reporting lower task costs for both than for the named competitors in its comparison. That could matter to organizations running repeated model-assisted work, where usage costs accumulate. The task-cost figures are benchmark-specific, however, and actual bills will depend on prompt length, output volume, processing tier and account limits.
The safety findings complicate the performance story. OpenAI’s system card places both models at its Critical cybersecurity threshold, the higher of its two stated risk levels. The company says more permissive cyber access requires extra vetting. It also says Astra is its first externally released model to meet its High threshold for AI self-improvement. Those are OpenAI’s risk assessments, not independent determinations that either model has caused harm.
Artificial Analysis found no improvement in its Coding Agent Index and reported hallucination rates of 59% for Astra and 55% for Sol at maximum effort. Those results mean a high overall index score does not guarantee better performance in every work category. Teams evaluating the release will need to weigh price and benchmark strengths against task-specific reliability and safety controls.
Products Worth Considering

Amazon Kindle Scribe Colorsoft 64GB with Premium Pen

Bose Portable Smart Speaker with Alexa and Google Assistant, Black

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)
As an affiliate, we earn on qualifying purchases.
What Changes Across the Model Line
OpenAI describes Astra as its model for the most demanding projects and Sol as the lower-cost choice for complex work. Both models accept text and images and generate text. Each has a 1,050,000-token context window, with up to 922,000 input tokens and 128,000 output tokens. Astra’s knowledge cutoff is May 18, 2026; Sol’s is June 1, 2026. Fine-tuning is not supported, according to the listed model information.
The third release, GPT-6.2 Astra for Law, appears on OpenAI’s pricing page as a legal model configuration with the ID gpt-6.2-astra-law. The source material says the available pages do not explain how it differs from standard Astra. Artificial Analysis had not assigned it an Intelligence Index score, although it had tested the legal configuration on one legal benchmark.
Pricing is not a single flat rate. Requests above 272,000 input tokens are charged at double the input and cached-input rates and 1.5 times the output rate for the whole request. Batch processing is half the Standard rate, while Fast mode is twice the rate. Ultrafast is listed at three times Standard through Feb. 28, 2027, and six times from March 1. Eligible background requests cost 30% less with a two-hour completion window or 40% less with an eight-hour window.
OpenAI’s developer guidance also describes a behavior change: starting with GPT-6.2, editing conversation history mid-conversation drops the model’s earlier reasoning. The company warns that answer quality can decline and the model may use more tokens repeating earlier work. API requests setting reasoning effort to “none” are rejected. These details may affect applications that revise a conversation rather than starting a fresh request.
Limits of the Launch-Day Evidence
The available information does not establish how Astra for Law differs from standard Astra, and it had no overall Intelligence Index score when the source was checked. A result on one legal benchmark cannot establish how it will perform across legal tasks or professional settings.
Artificial Analysis’s scores and task costs are a snapshot accessed on Nov. 17, 2026. The provided material does not include the full test-by-test results, so the specific areas where Astra leads or trails cannot be reported in detail here. The reported hallucination rates and coding-agent result also apply to the tests and configurations named by the source, not every real-world use.
OpenAI’s launch post was not accessible to the source publisher, leaving ChatGPT plan availability unclear. The material also does not specify how its Critical cybersecurity rating translates into restrictions for every customer, beyond the statement that more permissive access requires extra vetting. Independent evidence of real-world security outcomes was not provided.
Access, Pricing and Further Tests
Developers can check OpenAI’s model and pricing pages for current API availability and account-specific rate limits. The listed limits vary by usage tier, from 5,000 requests and 1 million tokens per minute at the lower listed tier to 15,000 requests and 40 million tokens per minute at the upper tier. Organizations considering deployment should also verify eligibility for Zero Data Retention and data residency options: the source says approved organizations can use US and EU residency with Standard, Fast and Batch processing, while Ultrafast is limited to US residency.
Further independent testing and clearer documentation will help establish how Astra for Law compares with standard Astra and whether the launch-day benchmark pattern holds across specific workflows. Buyers should confirm the applicable token rates and processing options before estimating costs, particularly for long prompts and Ultrafast use. OpenAI’s stated promotional period for legal pricing runs at least through Dec. 31, 2026; the source does not provide a later end date.
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, and $2 and $10, respectively, for Sol. Cached input costs $0.40 per million tokens for Astra and $0.10 for Sol. Long prompts and faster processing tiers can change the bill.
Which model scored higher in the independent index?
Astra scored 62 at Max and Sol scored 57 on Artificial Analysis’s Intelligence Index. Claude Fable 5.2 led the family comparison at 65; the more precise scores were 62.14 for Astra Max and 62.22 for Fable 5.2 Xhigh.
Does Astra improve coding-agent performance?
Not in the reported comparison. Artificial Analysis said there was no improvement in its Coding Agent Index. That finding applies to the index’s evaluation, not every coding task or deployment.
What does OpenAI’s Critical cybersecurity rating mean?
OpenAI’s system card rates both models Critical under its cybersecurity risk thresholds and says more permissive cyber access requires extra vetting. The source material does not detail the controls for every type of account or use.
Is GPT-6.2 Astra for Law a separate model?
OpenAI lists Astra for Law 6.2 as a legal model configuration, with its own model ID and the same listed standard rates as Astra. The available pages described in the source do not explain how it differs from standard Astra.
Source: ThorstenMeyerAI.com
