🔍 Read the full analysis: How Much Do GPT-6.2 Astra And Sol Cost, And How Do They Score? 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. Artificial Analysis scores Astra 62 and Sol 57, while its listed index task costs are $4.00 and $1.01, respectively; OpenAI also rates both models at its Critical cybersecurity risk threshold.
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on Nov. 17, alongside a legal configuration called Astra for Law 6.2. Same-day measurements by Artificial Analysis put Astra at 62 points and Sol at 57 on its Intelligence Index, while OpenAI’s system card rates both models at its Critical cybersecurity risk threshold. The launch gives developers new options at different price and capability levels, but the test results are mixed rather than a clear lead across the board.
OpenAI describes Astra as its model for the most demanding projects and Sol as a lower-cost option for complex everyday work. The company’s developer guide recommends Astra for “the most demanding tasks” and Sol for “complex work at lower cost.” Both models handle 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, according to the source material.
At standard short-context API rates, Astra costs $8 per million input tokens and $40 per million output tokens. Sol costs $2 and $10, respectively. Astra’s listed rates are 20% below GPT-6 Astra’s previous $10 input and $50 output prices; Sol retains the $2 and $10 rates. Cached input costs $0.40 per million tokens for Astra and $0.10 for Sol. Astra for Law is listed at Astra’s rates, though its promotional pricing is stated to be available at least through Dec. 31, 2026.
Artificial Analysis reports that one of its index tasks costs $4.00 with Astra at Max and $1.01 with Sol at Max. For comparison, 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 index’s task, not a universal price per user request; actual bills vary with token use and processing settings.
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
What the Scores and Prices Show
The results place the models among strong competitors, but the ranking depends on configuration and rounding. Artificial Analysis gives Astra a rounded Index score of 62 and Sol 57. Claude Fable 5.2 leads the model-family comparison at 65. Astra’s unrounded score at Max is 62.14, just below Fable 5.2 at Xhigh, at 62.22; Astra’s model page consequently ranks it third among 91 configurations, even though the source describes Astra as second among model families.
The task-cost figures may matter to teams running repeated evaluations or agent workflows: in this specific index comparison, Astra’s task costs less than those of the named rivals, and Sol’s costs less still. But low benchmark-task cost does not by itself establish lower total operating costs. Long prompts, output volume, speed tier, and background processing all affect the bill. Artificial Analysis also reports no improvement in its Coding Agent Index, and gives hallucination rates at Max effort of 59% for Astra and 55% for Sol. Those findings temper the headline scores and should be read as results from the stated evaluations, not guarantees about every deployment.
Safety is another practical factor for organizations considering access. OpenAI’s own system card places both models at its Critical cybersecurity threshold, the higher of the two risk thresholds described in the source. The card also calls Astra the first model OpenAI has released externally that meets its High threshold for AI self-improvement. OpenAI says more permissive cyber access requires extra vetting. The company’s risk ratings are its own assessments, rather than an independent safety certification.
Products Worth Considering

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

Amazon Kindle Scribe Colorsoft 64GB with Premium Pen

Samsung 27-Inch Odyssey G5 QHD Gaming Monitor (G50F)
As an affiliate, we earn on qualifying purchases.
How the Launch Fits OpenAI’s Lineup
The release separates OpenAI’s offering into a higher-capability model and a lower-priced option. Astra’s price cut compared with GPT-6 Astra applies to the listed standard rates: input and output each fall by 20%, while cached input is also cheaper. Sol remains at the same listed input and output prices as before. The developer guide presents the choice as a trade-off between the most demanding work and lower-cost complex tasks.
Costs can change substantially with usage. Requests containing more than 272,000 input tokens are charged at twice the short-context input and cache rates and 1.5 times the output rate for the whole request, according to the pricing details supplied. Batch processing is half the Standard rate; Fast is twice Standard. Ultrafast is 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 or 40% for an eight-hour window.
OpenAI also lists Astra for Law 6.2, model ID gpt-6.2-astra-law, as a legal configuration. The available pages do not describe how it differs from Astra. Artificial Analysis has not assigned it an Intelligence Index score, although it has tested it on one legal benchmark. OpenAI’s launch post was inaccessible to the source author, so the ChatGPT plans that include the models could not be confirmed.
Open Questions on Access and Testing
The available information does not confirm which ChatGPT plans include Astra or Sol, because OpenAI’s launch post could not be accessed by the source author. It also does not explain what technical or product differences distinguish Astra for Law from the standard Astra model. Its performance on the broader Intelligence Index remains unscored.
Benchmark scores describe defined tests and settings, not every real-world task. Artificial Analysis reports the models’ index scores, task costs, hallucination rates, and coding-agent result, but the supplied material does not provide all test-by-test results or enough detail to generalize these measurements to every workload. The Critical cybersecurity rating is OpenAI’s own assessment; the source does not provide an independent evaluation of that risk rating. Account-specific rate limits also vary by usage tier.
What Developers Should Check Next
Developers evaluating the models can compare them on their own workloads, paying attention to output quality, token consumption, latency, and the selected processing tier—not just the headline Index score. They should also check their account’s current rate limits and access eligibility, particularly for cyber-related use, where OpenAI says more permissive access requires additional vetting.
Further details may clarify ChatGPT availability and the differences in the legal configuration. Teams using the API should account for a behavior change noted by OpenAI: starting with GPT-6.2, editing conversation history mid-conversation drops the model’s earlier reasoning, which the company says can reduce answer quality and lead to repeated work and more token use. OpenAI also rejects API requests that set reasoning effort to “none.” The legal model’s promotional pricing is listed through at least Dec. 31, 2026, while Ultrafast pricing is scheduled to rise on March 1, 2027.
Key Questions
How much does GPT-6.2 Astra cost?
At standard short-context API rates, Astra costs $8 per million input tokens and $40 per million output tokens. Cached input is $0.40 per million tokens. Long prompts and faster processing tiers can cost more.
How much does GPT-6.2 Sol cost?
Sol costs $2 per million input tokens and $10 per million output tokens at standard short-context rates. Cached input costs $0.10 per million tokens.
What scores did Astra and Sol receive?
Artificial Analysis reports rounded Intelligence Index scores of 62 for Astra at Max and 57 for Sol at Max. Astra’s unrounded score is 62.14; the leading configuration in the cited comparison, Claude Fable 5.2 at Xhigh, scores 62.22.
Are Astra and Sol proven to be better at coding?
No such conclusion is supported by the supplied results. Artificial Analysis reports no improvement in its Coding Agent Index. Performance can vary by coding task and configuration.
What does OpenAI’s Critical cybersecurity rating mean?
OpenAI’s system card places both models at its Critical cybersecurity risk threshold, the higher of the two thresholds described in the source. OpenAI says more permissive cyber access requires extra vetting; the supplied material does not include an independent assessment of the rating.
Source: ThorstenMeyerAI.com
