AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What’s Behind The GPT-6.2 Astra And Sol Scores And Pricing? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys 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 Sol on November 17, 2026, alongside a legal configuration called Astra for Law 6.2. Artificial Analysis ranked Astra near the top of its Intelligence Index and reported lower costs per benchmark task than several rivals, while also finding no improvement in its Coding Agent Index. OpenAI’s same-day safety report rates both models Critical for cybersecurity; some product details, including ChatGPT availability and how the legal configuration differs, remain unclear.

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, alongside a legal configuration called Astra for Law 6.2. Independent testing placed Astra among the top-performing model families and found lower costs per benchmark task than several competitors, but results were mixed: the Coding Agent Index showed no improvement, and OpenAI rated both new models Critical for cybersecurity risk.

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 with 65. Astra’s more precise score was 62.14, just below Fable 5.2 Xhigh’s 62.22; Astra’s model page ranked that configuration third among 91. Artificial Analysis said Astra gained five points over GPT-6 Astra, while Sol gained two and scored about the same as the earlier Astra at a substantially lower benchmark-task cost.

On the index’s task-cost measure, a task cost $4.00 with Astra Max and $1.01 with Sol Max. The corresponding reported costs were $26.54 for Claude Fable 5.2 Max and $5.48 for Gemini 4 Carbon High. These are costs for the index’s benchmark task, not a universal price for a customer’s workload. OpenAI’s listed standard API rates are $8 per million input tokens and $40 per million output tokens for Astra, compared with $2 and $10 for Sol. Astra’s rates are 20% below the previous GPT-6 Astra’s listed $10 input and $50 output rates.

OpenAI describes Astra as its model for demanding projects and Sol as its lower-cost option for complex work and agents. Both accept text and images, have a 1,050,000-token context window, and can return up to 128,000 output tokens. Astra for Law appears in OpenAI’s pricing information at Astra’s standard listed rates, but the available pages do not explain how it differs from the general model. Artificial Analysis had not published an Intelligence Index score for the legal configuration at the time of its November 17 review.

At a glance
reportWhen: Released November 17, 2026; benchmark a…
The developmentOpenAI released GPT-6.2 Astra, GPT-6.2 Sol and Astra for Law 6.2, with independent benchmark results and pricing published on November 17, 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.

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

Strong Scores, Uneven Gains

The release gives developers two models with different price and capability positions, and the published benchmark figures suggest the cheaper Sol may offer a lower-cost route to results near those of the previous Astra on this particular index. Astra’s high score also puts it close to the index leader, though the gap between the two reported top scores is narrow and the ranking depends on the configuration being compared.

The results do not establish that either model is best for every task. Artificial Analysis reported no improvement in its Coding Agent Index and hallucination rates of 59% for Astra and 55% for Sol at maximum effort. Those measures point to limits that may matter for teams evaluating coding or accuracy-sensitive applications. Benchmark performance and cost can also differ from results on an organization’s own prompts, workloads and settings.

Safety findings add another consideration for deployment. OpenAI’s system card rates both models Critical for cybersecurity, its higher of two stated risk thresholds, and says additional vetting is needed for more permissive cyber access. The card also identifies Astra as the first OpenAI model released externally to meet its High threshold for AI self-improvement. These are OpenAI’s risk assessments, not independent confirmation that a model will cause harm in a specific use.

Products Worth Considering

Apple MacBook Neo 13-inch Laptop with A18 Pro Chip, 8GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Blush

Apple MacBook Neo 13-inch Laptop with A18 Pro Chip, 8GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Blush

As an affiliate, we earn on qualifying purchases.

Prices Depend on How You Use Them

The listed token prices are standard short-context rates, not a complete estimate for every API request. OpenAI says requests using more than 272,000 input tokens are billed at twice the input and cached-input rates and 1.5 times the output rate for the entire request. Batch processing costs half the Standard rate, while Fast costs twice as much. Ultrafast is listed at three times Standard through February 28, 2027, then six times Standard beginning March 1, 2027.

Eligible background requests receive a stated discount: 30% below Standard for a two-hour completion window or 40% below for an eight-hour window. OpenAI says promotional pricing for Astra for Law is available at least through December 31, 2026. Rate limits vary by usage tier; the listed tiers range from 5,000 requests and 1 million tokens per minute to 15,000 requests and 40 million tokens per minute. Approved organizations can use Zero Data Retention and US or EU data residency with Standard, Fast and Batch processing; Ultrafast is listed with US residency only.

OpenAI’s developer guide also describes an API behavior change: if a conversation history is edited mid-conversation, the API drops the model’s earlier reasoning. OpenAI warns that answer quality can decline and the model may spend more tokens repeating work. Requests that set reasoning effort to “none” are rejected. The source article reports that OpenAI’s launch post was inaccessible, so it does not establish which ChatGPT plans include the models.

Open Questions on Access and Safety

The available information does not establish which ChatGPT plans include Astra or Sol, or when those plans may receive them. OpenAI’s pricing page lists Astra for Law but does not describe the configuration’s differences from Astra. Artificial Analysis had not given it an Intelligence Index score, although it had tested the legal configuration on one legal benchmark.

The reported benchmark scores and task costs cover specified configurations and tests, and should not be read as guaranteed outcomes for other prompts or production workloads. OpenAI’s system card supplies the cybersecurity and self-improvement ratings, but the source material does not provide enough detail to independently assess the underlying evaluations or explain the practical access restrictions for each customer. The source article also reports hallucination rates at maximum effort without enough accompanying detail here to generalize those figures across settings.

Developers Can Test Real Workloads

The next useful comparison will come from testing Astra and Sol on the tasks developers actually plan to run, including coding, agent workflows and accuracy-sensitive work. Teams calculating costs should check context length, processing tier, background eligibility and usage limits against OpenAI’s current pricing, rather than rely only on the benchmark-task figures.

Further clarification from OpenAI could establish ChatGPT availability, the practical differences in Astra for Law, and how its safety assessment affects access to cybersecurity capabilities. The announced Ultrafast pricing schedule also changes on March 1, 2027, while the legal-model promotion is stated to last at least through December 31, 2026. These dates may affect cost comparisons, but do not change the reported benchmark results.

Key Questions

What did OpenAI release on November 17, 2026?

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, plus Astra for Law 6.2, a legal model configuration.

How did Astra and Sol score?

Artificial Analysis reported an Intelligence Index score of 62 for Astra Max and 57 for Sol Max. Claude Fable 5.2 led the model-family comparison with 65; Astra’s precise 62.14 score was just below Fable 5.2 Xhigh’s 62.22.

How much do the models cost through the API?

OpenAI lists standard short-context rates of $8 per million input tokens and $40 per million output tokens for Astra, and $2 and $10 for Sol. Long prompts and processing tiers can change the bill.

Are the new models better at coding?

The reported results do not show a general coding improvement: Artificial Analysis said there was no improvement in its Coding Agent Index. Performance may differ across coding tasks and configurations.

What did OpenAI say about cybersecurity risk?

OpenAI’s system card rates both models Critical for cybersecurity and says more permissive cyber access requires extra vetting. This is the company’s assessment; the available material does not detail all access rules.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Greenland: US Oil Company Brings Equipment Ashore – Nuuk Protests

A US-based oil firm has brought equipment ashore in Greenland, prompting protests from Nuuk. The development raises environmental and geopolitical concerns.

Why ByteDance’s Cautious AI Rollout Is A Model For Industry Success

ByteDance Seed describes its AI approach as ‘slow first, fast afterwards,’ potentially shaping industry standards. Impact remains unverified.

The Real Cost Of A Local-Inference Rig In 2026

Analyzing the hardware costs for local AI inference in 2026, including VRAM limits, hardware tiers, and value strategies for owning models.

AI And Ergonomics Combined: 14 Top Office Chairs For 2026

Discover the top 13 office chairs for 2026 combining AI insights and ergonomic design, enhancing comfort and productivity in modern workspaces.