🔍 Read the full analysis: GPT-6.2 Astra And Sol: The Numbers And Workplace Shifts To Know on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI released GPT-6.2 Astra and Sol on Nov. 17, 2026, positioning Astra for demanding work and Sol for lower-cost everyday tasks. Artificial Analysis placed Astra near the top of its index, while reporting no gain in its Coding Agent Index; OpenAI rated both models Critical for cybersecurity. Pricing is lower than for GPT-6 Astra, but real costs vary by processing tier and prompt length.
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, presenting Astra as its model for demanding projects and Sol as a lower-cost option for everyday work. Same-day independent tests put Astra near the top of Artificial Analysis’s model-family index, while OpenAI’s safety report rated both models Critical for cybersecurity under its risk framework.
Artificial Analysis gave Astra a score of 62 at its Max setting and Sol a score of 57 on its Intelligence Index. The index leader, Anthropic’s Claude Fable 5.2, scored 65 at the family level. The gap between the top systems is narrower in the unrounded figures: Astra scored 62.14 and Fable 5.2 scored 62.22 in the cited configurations. Astra’s model page lists it third among 91 configurations, a different ranking from the family-level comparison.
The evaluation also found a cost advantage on one index task: $4.00 for Astra Max and $1.01 for Sol Max, compared with $26.54 for Claude Fable 5.2 Max and $5.48 for Gemini 4 Carbon High. These are task costs in Artificial Analysis’s test, not general costs for every workload. The index reported no improvement in the Coding Agent Index and gave maximum-effort hallucination rates of 59% for Astra and 55% for Sol.
OpenAI lists standard API prices of $8 per million input tokens and $40 per million output tokens for Astra, down 20% from GPT-6 Astra. Sol is priced at $2 per million input tokens and $10 per million output tokens. Astra for Law 6.2 is listed separately at Astra’s standard rates, but the available product pages do not explain how its capabilities differ.
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
Performance Comes With Safety Trade-offs
The launch gives developers a new choice between higher capability and lower per-task cost, but the evaluation does not show a straightforward improvement in every area. Astra’s close score to the index leader and lower measured task cost may matter to organizations running complex workflows. Sol’s score and cost may appeal to teams with a larger volume of less demanding tasks.
Those comparisons need to be weighed against the cybersecurity risk rating. OpenAI’s system card says both models meet its Critical threshold for cybersecurity, the higher of the two risk thresholds in the company’s framework. OpenAI says more permissive cyber access requires additional vetting. The rating is a company assessment, not independent verification that the models have caused harm; it signals that OpenAI considers their capabilities relevant to cyber risk.
Costs also depend on how a model is used. Requests above 272,000 input tokens incur higher rates across the request, while Fast and Ultrafast processing cost more than Standard. Organizations comparing the headline price with their likely bills will need to account for prompt sizes, processing speed, and any eligible background-work discounts.
Popular right now

TCL 55-Inch F7D Series QLED 4K UHD Smart Fire TV (2026)

Dyson Airstrait Styling Brush

Roborock Q10 S5+ Robot Vacuum and Mop with Self-Emptying Station
As an affiliate, we earn on qualifying purchases.
How the New Models Are Positioned
OpenAI’s developer guide describes Astra as suited to the most demanding tasks and Sol as intended for complex work at lower cost. Both accept text and images, generate text, and have a listed context window of 1,050,000 tokens, with up to 922,000 input tokens and 128,000 output tokens. Their stated knowledge cutoffs are May 18, 2026, for Astra and June 1, 2026, for Sol. OpenAI says fine-tuning is not supported.
Artificial Analysis said Astra gained five points over GPT-6 Astra and Sol gained two points, reaching a score it described as on par with GPT-6 Astra at about 15% of the cost per task. The same evaluation found Astra leading some tests and trailing on others. The source material identifies a lead in agentic science workflows, but does not provide enough of the test-by-test results to characterize every comparison.
OpenAI also documented an API behavior change: when conversation history is edited mid-conversation, GPT-6.2 drops its earlier reasoning. The company warns that answer quality may decline and the model may spend more tokens repeating prior work. Requests that set reasoning effort to “none” are rejected. The launch article was not accessible to the source publisher, so its ChatGPT plan availability could not be established.
Open Questions on Access and Risk
The available information does not establish which ChatGPT plans include Astra or Sol, because the launch post was inaccessible to the source publisher. OpenAI’s listing for Astra for Law gives a model ID and price but does not describe how that configuration differs from standard Astra. Artificial Analysis had not assigned the legal model an Intelligence Index score, although it had tested it on one legal benchmark.
The index and task-cost figures reflect specific tests and model settings, not every real-world use. The supplied material does not give a full breakdown of the benchmark results, so the exact size of Astra’s advantages and disadvantages across all tests is not clear. OpenAI’s Critical cybersecurity rating is its own safety assessment; details about the additional vetting for more permissive access were not provided.
Pricing terms may also change a deployment’s cost. Rates for requests with more than 272,000 input tokens rise for the whole request; Ultrafast processing is listed at three times Standard through February 28, 2027, and six times Standard starting March 1, 2027. The source says promotional pricing for Astra for Law is available at least through December 31, 2026, but does not establish what will follow.
What Developers Should Track
Developers can assess the models through OpenAI’s API, subject to account-tier limits and the listed pricing. Before adopting either model, teams will need to test their own workloads for quality, latency and total cost, especially where long prompts or faster processing are required. They should also account for the documented behavior when conversation history is edited.
Further details could clarify Astra for Law’s distinct capabilities, its benchmark performance, and which ChatGPT plans include the new models. Organizations considering cyber-related use will also need to follow OpenAI’s access and vetting requirements. The next useful evidence will be independent testing across more tasks and real deployment results; the material available on launch day does not yet supply either.
Key Questions
When did OpenAI release GPT-6.2 Astra and Sol?
OpenAI released both models on November 17, 2026. Astra for Law 6.2 was also listed that day as a legal model configuration.
Which model scored higher in the reported index?
Astra scored 62 at its Max setting and Sol scored 57 on Artificial Analysis’s Intelligence Index. Claude Fable 5.2 led the family comparison at 65. Astra’s unrounded score was 62.14, just below Fable 5.2’s 62.22 in the cited configurations.
How much do Astra and Sol cost through the API?
Standard prices are $8 per million input tokens and $40 per million output tokens for Astra, and $2 per million input tokens and $10 per million output tokens for Sol. Long prompts and faster processing tiers can raise the bill.
What does OpenAI’s Critical cybersecurity rating mean?
OpenAI’s system card says both models meet its Critical threshold for cybersecurity, the higher of the two risk thresholds it uses. The company says more permissive cyber access requires extra vetting. This is OpenAI’s assessment, not a finding that the models have caused harm.
Are GPT-6.2 Astra and Sol available in ChatGPT?
The source material does not confirm which ChatGPT plans include the models. It says OpenAI’s launch post was inaccessible to the publisher, so availability through ChatGPT remains unclear.
Source: ThorstenMeyerAI.com
