GPT-6.2 Astra And Sol: What OpenAI’s New Models Cost, Score And Change For Work
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: GPT-6.2 Astra And Sol: What OpenAI’s New Models Cost, Score And Change For Work on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
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, alongside a legal configuration called Astra for Law. Independent testing puts Astra near the top of the Artificial Analysis Intelligence Index and reports lower costs per benchmark task than several rivals, but finds no improvement in its Coding Agent Index. OpenAI rates both models Critical for cybersecurity and reports other safety regressions.

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. Same-day independent tests place Astra among the leading model families, while OpenAI’s safety report rates both models Critical for cybersecurity—a combination of performance gains, cost reductions and elevated risks that matters to developers weighing deployment.

OpenAI describes Astra as its model for “our most capable model for your hardest projects, from start to finish,” and Sol as its “go-to model for everyday work, from coding to powering agents at scale.” A third product, Astra for Law 6.2, is listed as a legal configuration of Astra. The material available for this report does not explain how the legal configuration differs from the standard model.

OpenAI lists standard API prices of $8 per million input tokens and $40 per million output tokens for Astra, down from $10 and $50 for GPT-6 Astra. Sol is priced at $2 per million input tokens and $10 per million output tokens, the same rates reported for its predecessor. Cached input costs $0.40 per million tokens for Astra and $0.10 for Sol. Astra for Law is listed at Astra’s standard rates, with promotional legal pricing available at least through Dec. 31, 2026, according to OpenAI.

Artificial Analysis’s Intelligence Index gives Astra a score of 62 at its Max setting and Sol 57. The index’s reported leader, Claude Fable 5.2, scores 65; however, Astra’s unrounded score of 62.14 is just below Fable’s 62.22 at Xhigh, and Astra’s model page ranks it third among 91 configurations. Artificial Analysis reports a benchmark-task cost of $4 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.

At a glance
announcementWhen: Announced November 17, 2026; benchmark…
The developmentOpenAI launched GPT-6.2 Astra and Sol, with same-day independent benchmark results, pricing information and a safety report.

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

Lower Task Costs, Higher Cyber Risk

The launch gives organizations a new option to compare not just on headline benchmark scores, but also on the cost of running particular workloads. In Artificial Analysis’s measured task, Astra and Sol cost less than the named rival configurations. That may matter to teams running agents or repeated model evaluations, though the result is specific to the index’s task and is not a general estimate of total operating costs.

Performance is not uniformly better. Artificial Analysis says GPT-6.2 Astra shows no improvement in its Coding Agent Index, while reporting hallucination rates of 59% for Astra and 55% for Sol at maximum effort. Those measurements make it important for customers to test the models on their own tasks rather than infer that a high overall index score means better results in every use case.

OpenAI’s cybersecurity rating adds a deployment concern. The company says both models meet its Critical risk threshold, the higher of the two thresholds used in its framework, and says access to more permissive cyber capabilities now requires additional vetting. The release therefore combines lower benchmark-task costs with a need for careful controls, particularly in cybersecurity-sensitive settings.

Products Worth Considering

Jackery HomePower 3600 Plus Portable Power Station, 3584Wh, 3600W

Jackery HomePower 3600 Plus Portable Power Station, 3584Wh, 3600W

DJI Mini 4K Drone Combo with 4K Camera, 3-Axis Gimbal, and Two Batteries

DJI Mini 4K Drone Combo with 4K Camera, 3-Axis Gimbal, and Two Batteries

As an affiliate, we earn on qualifying purchases.

How the Models Differ

OpenAI’s developer guide recommends Astra for the most demanding tasks and Sol for complex work at lower cost. Both models accept text and images and generate text. Each has a 1,050,000-token context window, accepts up to 922,000 input tokens and can return up to 128,000 output tokens. OpenAI lists knowledge cutoffs of May 18, 2026, for Astra and June 1, 2026, for Sol; fine-tuning is not supported.

Pricing can rise above the standard rates. For requests exceeding 272,000 input tokens, OpenAI says input and cached-input rates double, while output is charged at 1.5 times the short-context rate for the whole request. Batch processing costs half the standard rate; Fast mode costs twice as much. Ultrafast costs three times the standard rate through Feb. 28, 2027, and six times from March 1. Eligible background requests are discounted by 30% for a two-hour completion window or 40% for an eight-hour window.

Artificial Analysis tests models across nine evaluations and reports scores, task costs and completion times. Its results describe Astra as trailing only Anthropic’s Claude Fable 5.2 at the model-family level, while the precise ranking varies by configuration and score. OpenAI also published a system card on launch day. The source material notes that the article supplying these details was AI-assisted and that its author disclosed a conflict because Anthropic competes with OpenAI; neither company nor Artificial Analysis reviewed or sponsored it.

Open Questions on Access and Testing

OpenAI’s launch post was not accessible to the source report, so it could not confirm which ChatGPT plans include the new models or the consumer-facing availability details. The available material also does not establish how Astra for Law differs technically from standard Astra. Artificial Analysis had not assigned the legal configuration an Intelligence Index score, although it had tested it on one legal benchmark.

The benchmark figures are snapshots from tests accessed on Nov. 17, 2026, not guarantees of performance on every workload. The source material provides only part of the test-by-test comparison and does not specify every task result in the supplied excerpt. OpenAI’s safety report lists regressions alongside improvements, but the supplied information does not detail each one. The full scope of the Critical cyber classification and the conditions for individual organizations’ access remain unclear here.

Costs also depend on prompt length, processing speed, background settings and account limits. The reported cost per benchmark task should not be read as a universal cost per job or as a forecast of a customer’s bill. Actual expenses will depend on token use, selected service tier and workload.

Deployment Terms and Evaluations

Organizations considering the models can compare Astra and Sol against their own tasks, including coding, agent workflows and any legal work for which Astra for Law is being evaluated. They will need to account for long-context surcharges and speed-tier pricing, as well as the limits imposed by their API usage tier. OpenAI lists rate limits starting at 5,000 requests and 1 million tokens per minute, rising on listed tiers 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. Further clarification on ChatGPT plan access, the legal configuration and cyber-vetting requirements would help customers assess availability and governance. Additional independent testing, including more results for the legal configuration and real-world evaluations, will show how the launch’s benchmark gains and safety findings apply beyond the initial measurements.

Key Questions

What did OpenAI release on November 17, 2026?

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, and listed Astra for Law 6.2 as a legal model configuration. The supplied information does not explain the technical differences between Astra and its legal configuration.

How much do GPT-6.2 Astra and Sol cost?

Standard API 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 requests and faster processing tiers can cost more; Batch and eligible background requests can cost less.

How did the models score in independent testing?

Artificial Analysis scored Astra at 62 at Max and Sol at 57 on its Intelligence Index. Claude Fable 5.2 led the cited model-family comparison with 65. Astra’s precise score and rank depend on configuration; its model page ranks Astra Max third among 91 configurations.

What safety concern did OpenAI report?

OpenAI’s system card rates both GPT-6.2 models Critical for cybersecurity, the higher of the two risk thresholds it uses. OpenAI says more permissive cyber access requires additional vetting; the supplied material does not provide all details of that process.

Are the new models available in ChatGPT?

The source material does not confirm which ChatGPT plans include Astra or Sol. It reports API pricing and developer information, but says OpenAI’s launch post could not be accessed for plan availability details.

Source: ThorstenMeyerAI.com

You May Also Like

A Week Of US Tech Outages Spans Banks, Flights, Email, And The NSA

Banks, flights, email and government systems went down in a week of outages. Organizations cite separate technical causes; no attack link is confirmed.

Inside The 2024 Funding Roundup For Li Auto And SenseTime-Spun Chip Firms

A eu.36kr.com headline reports ¥11.2 billion in 2024 funding for chip firms linked to Li Auto and SenseTime, but provides no deal breakdown.

With BioWare Dead, EA Lays Out Its ‘Next Chapter,’ And It’s Pretty Much As Predicted: Guns And Football Are Fine, And Everything Else Is Under The Microscope

EA says it will prioritize fewer large projects, including Battlefield and EA Sports, while confirming BioWare’s closure and pausing Project Lotus.

How Block Blast Harvests Its Players For Ad Money

A technical report says Block Blast downloads server-selected game modules, including traits named for difficulty, revival and ad strategy. Key claims remain unverified.