OpenAI’s GPT-6.2 Astra And Sol: A Guide To Work, Cost, And Scores
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: OpenAI’s GPT-6.2 Astra And Sol: A Guide To Work, Cost, And Scores on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • 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 November 17, 2026, alongside a legal configuration called Astra for Law. Artificial Analysis ranked Astra near the top of its Intelligence Index and found lower costs per test task than several rivals, while reporting no improvement in its Coding Agent Index; OpenAI rated both models Critical for cybersecurity.

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026, positioning Astra for demanding projects and Sol for lower-cost everyday work. Same-day independent tests put Astra close to the top of Artificial Analysis’s Intelligence Index and reported lower cost per benchmark task than several rivals, while OpenAI’s safety report rated both new models Critical for cybersecurity.

OpenAI describes Astra as its model for “your hardest projects, from start to finish” and Sol as its choice for everyday work, including coding and agents at scale. A third offering, GPT-6.2 Astra for Law, also appeared on the company’s pricing page as a legal configuration of Astra. The available pages do not explain how that configuration differs from the standard model.

Artificial Analysis scored Astra Max at 62 and Sol Max at 57 on its Intelligence Index, accessed November 17. Claude Fable 5.2 led with 65 when scores were rounded. The unrounded scores were closer: Astra scored 62.14 and Fable 5.2 scored 62.22. Artificial Analysis therefore lists Astra Max third among 91 configurations on its model page, even as its release summary describes OpenAI as trailing only the Claude model at the family level.

The benchmark provider’s estimated cost for one index task was $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. Those figures apply to the provider’s benchmark task, not every customer workload. Artificial Analysis also said the release brought “No improvement in Coding Agent Index” and reported hallucination rates of 59% for Astra and 55% for Sol at maximum effort.

At a glance
announcementWhen: Announced November 17, 2026; benchmark…
The developmentOpenAI launched GPT-6.2 Astra and Sol, with same-day independent benchmark results and a safety report showing both models meet the company’s Critical cybersecurity threshold.

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

Costs, Capability and Cyber Risk

The launch gives developers two models with different price and capability targets, but the headline benchmark scores do not settle which is the better choice for a particular application. OpenAI lists Astra at $8 per million input tokens and $40 per million output tokens; Sol costs $2 and $10, respectively. Astra’s listed rates are 20% below those for GPT-6 Astra, while Sol remains at $2 input and $10 output.

Artificial Analysis’s task-cost estimates suggest that the models may be less expensive on its test suite than some competing systems. Actual bills can differ with prompt length, usage tier, processing speed and task design. OpenAI says that requests exceeding 272,000 input tokens have doubled input and cached-input rates and output rates 1.5 times the short-context price for the whole request. Batch processing costs half the standard rate; Fast costs twice as much. Ultrafast is priced at three times standard through February 28, 2027, then six times from March 1.

The safety designation is a separate concern for organizations considering deployment. OpenAI’s system card rates both models Critical for cybersecurity, its higher of two cyber-risk thresholds, and says more permissive cyber access requires extra vetting. The same report describes Astra as the first OpenAI model released externally that meets the company’s High threshold for AI self-improvement. These are OpenAI’s assessments, not independent confirmation of real-world outcomes.

Products Worth Considering

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

Bose QuietComfort Wireless Noise-Cancelling Headphones, Sandstone

Bose QuietComfort Wireless Noise-Cancelling Headphones, Sandstone

As an affiliate, we earn on qualifying purchases.

How the New Models Differ

Both models accept text and images and generate text. OpenAI lists 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 model information provided.

OpenAI’s developer guide recommends Astra for the most demanding work and Sol for complex work at lower cost. It also flags a change for API users: if conversation history is edited mid-conversation, GPT-6.2 drops the model’s earlier reasoning. OpenAI warns that answer quality can decline and the model may use additional tokens repeating earlier work. Requests that set reasoning effort to “none” are rejected.

The pricing page lists Astra for Law at the same standard token rates as Astra. OpenAI says promotional pricing for that legal configuration is available at least through December 31, 2026. Artificial Analysis had not assigned it an Intelligence Index score as of November 17, though it had tested the configuration on one legal benchmark.

What the Scores Do Not Show

The benchmark results cover a defined set of tests and do not establish how Astra or Sol will perform across every customer task. Artificial Analysis’s cost figures are estimates for one task in its index; the source material does not provide enough detail here to translate them into a typical user’s monthly bill. The reported hallucination rates are also tied to the provider’s test and maximum-effort settings, rather than a universal rate.

OpenAI’s launch post was not accessible to the source publication, so which ChatGPT plans include the models remains unconfirmed. The available information also does not explain Astra for Law’s technical or behavioral differences from Astra, or give a full result for its legal benchmark. OpenAI’s cyber and self-improvement ratings are company assessments; the material does not provide independent validation of those classifications or detailed evidence about how risks may vary by deployment.

Availability and Pricing to Watch

Developers can check OpenAI’s model and pricing pages for current API access, account-specific rate limits and data-handling options. Listed usage tiers range from 5,000 requests and 1 million tokens per minute to 15,000 requests and 40 million tokens per minute. OpenAI says 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.

Key points to watch include whether OpenAI clarifies access through ChatGPT plans, publishes more detail about Astra for Law, or updates its safety documentation and benchmark results. The legal model’s promotional pricing is stated to run at least through December 31, 2026, while the Ultrafast pricing schedule changes on March 1, 2027. Those dates and the provider’s follow-up results may affect deployment and cost decisions.

Key Questions

What did OpenAI release?

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. A legal configuration, GPT-6.2 Astra for Law, appeared on the same day.

Which model scored higher on Artificial Analysis?

Astra Max scored 62 and Sol Max scored 57 on the Intelligence Index. Claude Fable 5.2 led with a rounded score of 65; Astra’s unrounded score was 62.14, slightly below Fable 5.2’s 62.22.

How much do Astra and Sol cost through the API?

Standard short-context pricing is $8 per million input tokens and $40 per million output tokens for Astra, and $2 input and $10 output for Sol. Long prompts and processing options can change the rate.

What cybersecurity rating did the models receive?

OpenAI’s system card rates both Astra and Sol Critical for cybersecurity, the higher of the two cyber-risk thresholds it uses. OpenAI says access to more permissive cyber capabilities requires extra vetting.

Are the models available in ChatGPT?

The source material did not confirm which ChatGPT plans include Astra or Sol. It describes the API models and pricing, but says OpenAI’s launch post was not accessible to the source publication.

Source: ThorstenMeyerAI.com

You May Also Like

Claude Sonnet 5.5 Nearly Matches Opus 5.5 Benchmarks At Up To 30% Lower Cost

A headline says Sonnet 5.5 nearly matches Opus 5.5 benchmarks at up to 30% less per task, but provides no scores or pricing method.

The Menu: What Ten Answers Reveal

A comprehensive analysis of how ten jurisdictions respond to automation and AI, revealing patterns in income, capital, work, skills, and institutions.

Exploring The Security Failures That Allowed Claude Hacks, Says Anthropic

Anthropic admits security flaws linked to hacking incidents involving its Claude AI models, raising questions about AI safety and defenses.

Which AI Model Is Best For Enhancing Your Coding Workflow?

An analysis of top AI models for software development, highlighting their strengths, best use cases, and what developers should consider.