GPT-6.2 Sol and GPT-6.2 Astra Are Now on Risphai: Benchmarks, Pricing and Pro Mode
OpenAI's GPT-6.2 Sol and GPT-6.2 Astra are live on Risphai with Pro mode. See benchmarks, API and credit prices, Astra vs Sol vs Claude Fable 5.2 and how to use them.
By Risphai
GPT-6.2 Sol and GPT-6.2 Astra replace GPT-6.1 Sol and GPT-6.1 Terra on Risphai.
GPT-6.2 Sol and GPT-6.2 Astra are now on Risphai. OpenAI released both models on November 17, 2026: Astra for the hardest professional work and Sol for fast, everyday tasks. You can use them today for chat, reasoning, writing and coding, with four reasoning levels from Medium to Max and a new Pro mode for harder problems. GPT-6.2 Sol is on every paid plan, and GPT-6.2 Astra is on Premium.
GPT-6.2 replaces GPT-6.1 Sol and GPT-6.1 Terra on Risphai. OpenAI's pitch for this release is cost per task: according to OpenAI, Astra beats its predecessor GPT-6 Astra on professional, legal and science benchmarks while using fewer AI words, and Sol nearly matches Astra on coding at about a fifth of the cost. Independent tests from Artificial Analysis confirm most of the gains, with a few caveats we cover below.
A cheaper top model. OpenAI priced GPT-6.2 Astra 20% lower than GPT-6 Astra, and cached input is 60% cheaper. GPT-6.2 Sol keeps GPT-6.1 Sol's price.
Fewer AI words per task. On GDPval v2, OpenAI's internal test of consulting, finance and design projects, OpenAI says Astra scores 6.2 percentage points higher than GPT-6 Astra while writing 22% fewer output AI words. Sol improves 5.3 points over GPT-6.1 Sol.
Better professional documents. OpenAI trained Astra to follow the templates and conventions of fields such as law, finance and accounting. It says slides pick up layouts, colors and fonts from templates, and that Astra's writing is more direct, with less filler.
Stronger science work. On Terminal-Bench-Science 0.1, OpenAI reports Astra 9 percentage points above GPT-6 Astra at less than a third of the cost.
Sol nearly matches Astra on code. On FrontierCode 2.0, OpenAI says Sol trails Astra by just 0.5 percentage points at about a fifth of the cost per task.
Better 3D and web design. On AA-3D, where people compare two blinded interactive scenes, Astra ranks first with 1,318 Elo, ahead of Claude Fable 5.2 at 1,271 and Sol at 1,259.
Both models always reason before they answer, so on Risphai the lightest setting is Low thinking, which we use in chat and writing. OpenAI also classifies both models at its Critical cybersecurity capability threshold and has added stricter safeguards for risky security requests. These are OpenAI's own statements, and we haven't tested them ourselves.
GPT-6.2 Benchmarks
Independent results from Artificial Analysis, as reported by The Decoder and MarkTechPost. Higher is better.
These results come from OpenAI and from Artificial Analysis, as reported by The Decoder, MarkTechPost and IBTimes on November 17 and 18, 2026. We haven't run these benchmarks ourselves. The best score in each row is in bold.
Benchmark
What it tests
GPT-6.2 Astra
GPT-6.2 Sol
Claude Fable 5.2
Source
Artificial Analysis Intelligence Index
Nine tests of knowledge work, coding, science and general knowledge
62
57
65
Artificial Analysis
AA-3D
Interactive 3D scenes, judged blind by people (Elo)
1,318
1,259
1,271
Artificial Analysis
AA-Robotics v0.1
Robotics tasks
79
78
75
Artificial Analysis
AA-Omniscience hallucination rate
How often a model answers wrongly instead of admitting it doesn't know (lower is better)
59%
55%
Not listed
Artificial Analysis
Agents' Last Exam v1.1
Multi-step professional workflows
+5.5 points vs Fable
Not listed
Baseline
OpenAI
Harvey LAB-AA v1.1
Legal assignments from detailed briefs, pass rate
15.0%, more than twice Fable's
Not listed
Less than half of Astra's
OpenAI, Artificial Analysis
A few things to keep in mind when you read these numbers:
Fable 5.2 still leads the overall independent ranking. On the Artificial Analysis Intelligence Index, Claude Fable 5.2 is first with 65, ahead of Astra (62), Gemini 4 Carbon (60), Claude Opus 5.5 (59) and Sol (57). Artificial Analysis also places OpenAI's models on its cost-performance frontier, where no other model is both cheaper and better.
OpenAI's head-to-head with Fable used different settings. OpenAI compared Astra at xhigh reasoning with Fable 5.2 at max, and Fable's Agents' Last Exam score covered 113 of 122 tasks after the leaderboard excluded nine.
Coding agents didn't improve. Artificial Analysis reports that Astra's Coding Agent Index stayed level with GPT-6 Astra, and Sol's fell 3 points compared with GPT-6.1 Sol.
Hallucination stays high. Artificial Analysis measured hallucination rates of 59% for Astra and 55% for Sol on AA-Omniscience.
Benchmarks measure specific tasks, so results in your own work can differ. Always double-check important answers, especially about money, law or health.
GPT-6.2 Astra vs GPT-6.2 Sol vs Claude Fable 5.2
All three models are on Risphai. Here's how they compare:
GPT-6.2 Sol
GPT-6.2 Astra
Claude Fable 5.2
Best for
Everyday chat, reasoning, writing and coding
Hard professional, legal and science projects
The hardest reasoning and coding
Input credits, per million AI words
1,000
4,000
4,000
Output credits, per million AI words
5,000
20,000
20,000
Artificial Analysis Intelligence Index
57
62
65
Context window
1.05 million AI words
1.05 million AI words
1 million AI words
Pro mode
Yes, on Premium
Yes, on Premium
No
Zero Data Retention on Risphai
Yes
Yes
No
Plans
Every paid plan
Premium
Premium
Astra and Fable cost the same per AI word on Risphai, but OpenAI reports that Astra often needs fewer AI words to finish a task. A simple rule: use GPT-6.2 Sol by default, and switch to Astra or Fable when a problem is very hard or the first answer isn't good enough. You can switch between them in the same chat. Compare all specs on the GPT-6.2 Sol vs GPT-6.2 Astra and GPT-6.2 Astra vs Claude Fable 5.2 pages, or read more about Fable in Claude Fable 5.2 Is Now on Risphai.
GPT-6.2 Pro Mode
GPT-6.2 Sol Thinking Pro and GPT-6.2 Astra Thinking Pro run the models in OpenAI's Pro reasoning mode, which is meant for harder problems. On Risphai, Pro mode uses the same rates as the standard model, but it usually thinks longer, so a response typically takes more time and uses more credits. You can choose Pro mode for reasoning and coding with Medium, High, Extra High or Max on the Premium plan.
Where You Can Use GPT-6.2 on Risphai
Function
Model option
Thinking levels
Plans
Chat and writing
GPT-6.2 Sol Thinking (Low)
Low
Every paid plan
Chat and writing
GPT-6.2 Astra Thinking (Low)
Low
Premium
Reasoning and coding
GPT-6.2 Sol Thinking
Medium, High, Extra High, Max
Every paid plan
Reasoning and coding
GPT-6.2 Astra Thinking
Medium, High, Extra High, Max
Premium
Reasoning and coding
GPT-6.2 Sol Thinking Pro and GPT-6.2 Astra Thinking Pro
Medium, High, Extra High, Max
Premium
Risphai picks a model for each request based on the task and your quality preset. Since November 18, 2026, the presets use:
Request
Cost Efficient
Fast
Balanced
Quality
Extreme Quality
Chat
Other model
Other model
Kimi K3.1 Pro Thinking (Low)
Kimi K3.1 Pro Thinking (Low)
GPT-6.2 Sol Thinking (Low)
Reasoning
Other model
Other model
GPT-6.2 Sol Thinking (Medium)
GPT-6.2 Sol Thinking (High)
GPT-6.2 Sol Thinking (Max)
Writing
Other model
Other model
Kimi K3.1 Pro Thinking (Low)
Kimi K3.1 Pro Thinking (Low)
GPT-6.2 Sol Thinking (Low)
Coding
Other model
Other model
Other model
Other model
GPT-6.2 Sol Thinking (High)
The Free plan uses Cost Efficient. The other presets are available on Paid plans, and in Advanced mode you can choose GPT-6.2 yourself. If GPT-6.2 is unavailable or your balance doesn't cover a request, Risphai automatically uses another model so your request still finishes, and shows you which model answered.
Which Reasoning Level to Choose
Level
Where
Best for
Low
Chat and writing
Quick answers, everyday questions and writing that needs light thinking
Medium
Reasoning and coding
Most reasoning and coding tasks, at fewer credits per answer
High
Reasoning and coding
Harder math, logic and multi-step coding
Extra High
Reasoning and coding
Difficult problems where accuracy matters most
Max
Reasoning and coding
The hardest problems, when time and credits matter less
Higher levels let the model think longer before it answers. That usually improves results on hard problems, but takes more time and credits. Risphai shows a summary of the model's thinking while it works.
Prompts to Try
GPT-6.2 Astra is strongest on long professional work, and GPT-6.2 Sol on fast everyday tasks. These requests are a good start:
"Read these three quarterly reports and build a valuation table with the assumptions behind every number." (Astra, with your files)
"List every deadline, fee and termination right in this contract, with the clause each one comes from." (Astra, with your file)
"Find the bug in this function, explain it and give me a fixed version with tests." (Sol, with your code)
"Turn these meeting notes into a one-page project brief with goals, owners and risks." (Sol)
"What does this chart show, and is the conclusion in the caption supported by the data?" (Sol, with a screenshot)
"Build an interactive 3D solar system in a single HTML file." (Astra or Sol)
On Risphai, GPT-6.2 can also run Python to check its math, analyze your data or draw charts, and runnable HTML, CSS and JavaScript can be previewed right in the chat. AI can make mistakes, so check important results.
Choose a plan. Selecting models needs the Lite, Standard or Premium plan or Pay As You Go credits. GPT-6.2 Astra and Pro mode need Premium. See the Pricing page.
Switch to Advanced mode and open the model settings in the chat box.
Pick GPT-6.2. Choose GPT-6.2 Sol Thinking (Low) or GPT-6.2 Astra Thinking (Low) for chat or writing, or GPT-6.2 Sol Thinking or GPT-6.2 Astra Thinking for reasoning or coding and set the level. On Premium, you can also choose the Thinking Pro versions.
Ask your question. You can save your selection as your default and switch models at any time.
GPT-6.2 Pricing on Risphai
GPT-6.2 Sol
GPT-6.2 Astra
Input, per million AI words
1,000 credits
4,000 credits
Output, per million AI words (including thinking)
5,000 credits
20,000 credits
For example, a request with 10,000 input AI words and 2,000 output AI words uses about 20 credits with GPT-6.2 Sol and about 80 credits with GPT-6.2 Astra. Because thinking counts as output, harder questions and higher reasoning levels use more, and Pro mode usually thinks longer. Repeated parts of a conversation can be cached, which lowers the cost of long chats. Like every response, the total also includes the processing charge described in the Terms. The exact credits are shown for each response, and current prices are on the GPT-6.2 Sol and GPT-6.2 Astra model pages.
Risphai doesn't train AI models on your conversations. GPT-6.2 Sol and GPT-6.2 Astra are also available in Zero Data Retention mode at the same price. See the Privacy Policy for how each privacy mode works and which providers handle your requests.
What Happens to GPT-6.1 Sol and GPT-6.1 Terra
GPT-6.2 replaces GPT-6.1 Sol and GPT-6.1 Terra on Risphai. If you had saved a GPT-6.1 model as your default, Risphai now uses GPT-6.2 Sol instead, with Low thinking in chat and writing and at least Medium in reasoning and coding. Your earlier conversations and their answers stay as they are.
Try GPT-6.2
GPT-6.2 Sol and GPT-6.2 Astra are live on Risphai now. Choose a plan on the Pricing page, switch to Advanced mode and select GPT-6.2, or go Premium to use GPT-6.2 Astra and Pro mode. All changes in this release are in the v0.1.16 release notes.
OpenAI, ChatGPT and GPT are trademarks of OpenAI. Claude is a trademark of Anthropic. Risphai isn't affiliated with or endorsed by OpenAI or Anthropic. Model details and results are taken from OpenAI's public model pages and reports by The Decoder, MarkTechPost and IBTimes, as of November 18, 2026.
GPT-6.2 Sol & Astra Are Now Available on RisphaiWatch on YouTube
Frequently Asked Questions
What is GPT-6.2?
GPT-6.2 is a family of two AI models that OpenAI released on November 17, 2026. GPT-6.2 Astra is OpenAI's most capable model for hard, long projects in fields such as science, law and finance. GPT-6.2 Sol is OpenAI's model for fast, everyday work, from coding to writing. Both read text and images, write text and have a 1.05 million AI word context window.
Is GPT-6.2 available on Risphai?
Yes. Since November 18, 2026, you can choose GPT-6.2 Sol Thinking and GPT-6.2 Astra Thinking on Risphai. In chat and writing they use Low thinking. In reasoning and coding you can pick Medium, High, Extra High or Max, and each model also has a Pro mode. GPT-6.2 Sol is on every paid plan and with Pay As You Go credits, and GPT-6.2 Astra and the Pro mode need the Premium plan.
What is the difference between GPT-6.2 Astra and GPT-6.2 Sol?
Astra is the larger model for the hardest work, and Sol is the faster, cheaper model for everyday tasks. On Risphai, Astra costs four times as much as Sol per AI word: 4,000 instead of 1,000 credits per million input AI words and 20,000 instead of 5,000 credits per million output AI words. On the Artificial Analysis Intelligence Index, Astra scores 62 and Sol 57. On OpenAI's FrontierCode 2.0 coding benchmark, Sol comes within 0.5 percentage points of Astra at about a fifth of the cost per task, according to OpenAI.
Is GPT-6.2 Astra better than Claude Fable 5.2?
It depends on the benchmark. OpenAI reports that Astra scores 5.5 percentage points higher than Claude Fable 5.2 on Agents' Last Exam, at about 5% of Fable's estimated cost per task, and passes more than twice as many Harvey LAB-AA legal assignments. But OpenAI compared Astra at its xhigh setting with Fable at its max setting, and Fable's score excluded nine tasks. On the independent Artificial Analysis Intelligence Index, Claude Fable 5.2 ranks first with 65 and Astra second with 62. Both models are on Risphai's Premium plan, so you can try both on your own tasks.
How much does GPT-6.2 cost on Risphai?
GPT-6.2 Sol costs 1,000 credits per million input AI words and 5,000 credits per million output AI words, including thinking. GPT-6.2 Astra costs 4,000 and 20,000 credits. Prompts over 272,000 input AI words cost more. Each response also includes the processing charge, and you see the exact credits for every response.
What is GPT-6.2 Pro mode?
GPT-6.2 Sol Thinking Pro and GPT-6.2 Astra Thinking Pro run the models in OpenAI's Pro reasoning mode, which is meant for harder problems. On Risphai, Pro mode uses the same rates as the standard model, but it usually thinks longer, so a response typically takes more time and uses more credits. Pro mode is available for reasoning and coding on the Premium plan.
Which plan do I need for GPT-6.2 Astra?
GPT-6.2 Astra is available on the Premium plan. Premium also includes GPT Pro models, Claude Fable 5.2 and 20,000 credits every month. GPT-6.2 Sol is available on the Lite, Standard and Premium plans and with Pay As You Go credits.
Can I use GPT-6.2 for free?
Not on Risphai's Free plan. The Free plan includes 100 credits every month and uses the Cost Efficient preset, which doesn't use GPT-6.2. To choose GPT-6.2 Sol yourself, you need the Lite, Standard or Premium plan or Pay As You Go credits. GPT-6.2 Astra needs Premium.
What happened to GPT-6.1 Sol and GPT-6.1 Terra on Risphai?
GPT-6.2 Sol and GPT-6.2 Astra replace GPT-6.1 Sol and GPT-6.1 Terra on Risphai since November 18, 2026. If you had saved a GPT-6.1 model as your default, Risphai now uses GPT-6.2 Sol with a supported thinking level instead. Your earlier conversations stay as they are.
Which quality presets use GPT-6.2?
Since November 18, 2026, the Balanced preset uses GPT-6.2 Sol Thinking (Medium) for reasoning, the Quality preset uses GPT-6.2 Sol Thinking (High) and the Extreme Quality preset uses GPT-6.2 Sol Thinking (Max). Extreme Quality also uses GPT-6.2 Sol Thinking (Low) for chat and writing and GPT-6.2 Sol Thinking (High) for coding. Balanced and Quality use Kimi K3.1 Pro Thinking (Low) for chat and writing.
What is the context window of GPT-6.2?
OpenAI lists a context window of 1.05 million AI words for both GPT-6.2 Sol and GPT-6.2 Astra, with up to 922,000 input AI words and up to 128,000 output AI words.
What is the knowledge cutoff of GPT-6.2?
OpenAI lists June 1, 2026 as the knowledge cutoff for GPT-6.2 Sol and May 18, 2026 for GPT-6.2 Astra. On Risphai, both models can use web search when a question needs newer information.
Does GPT-6.2 work with Zero Data Retention on Risphai?
Yes. GPT-6.2 Sol and GPT-6.2 Astra are available in Risphai's Zero Data Retention mode at the same price. See the Privacy Policy for how each privacy mode works.
gpt-6.2gpt 6.2gpt-6.2 astragpt 6.2 astragpt-6.2 solgpt 6.2 solgpt-6.2 progpt-6.2 astra progpt-6.2 sol progpt-6.2 benchmarksgpt-6.2 pricinggpt-6.2 astra vs claude fable 5.2gpt-6.2 astra vs solopenai gpt-6.2openaichatgpt alternativeai models