Every few months, OpenAI drops a new model and the internet goes wild for a day. Most of the time, I read the announcement, nod, and move on. This time, I actually sat up. OpenAI just released GPT-6.2 Astra and GPT-6.2 Sol, and the headline isn’t just “smarter.” It’s smarter and cheaper, which almost never happens at the same time.

If you use ChatGPT for work, run a small business, or build apps with AI, this update could change how much you get done in a day and how much you pay for it. Let’s walk through what’s new, who it’s for, and what it costs.

GPT-6.2 Astra and Sol: What OpenAI Just Released

OpenAI launched two new models this week. GPT-6.2 Astra is its new top model for the hardest projects. GPT-6.2 Sol is the faster, more affordable model meant for everyday work.

The big promise for both is simple: better results using fewer words behind the scenes. AI models charge and count usage by “tokens,” which are small chunks of text. Because these models use fewer output tokens to finish a job, you can get more done before you hit your usage limits, and businesses pay less per task.

On top of that, Astra is cheaper to use than the model it replaces. According to OpenAI’s official announcement, Astra’s input and output prices are 20% lower than GPT-6 Astra, with a 60% discount on cached input. Sol keeps the same prices as GPT-6.1 Sol while getting smarter.

Key Facts at a Glance

DetailWhat You Need to Know
CompanyOpenAI
New modelsGPT-6.2 Astra (top model) and GPT-6.2 Sol (everyday model)
Who gets it in ChatGPTPlus, Pro, Enterprise and Edu users, in ChatGPT Work and Codex
Ultrafast modePro 500 on both models; Pro 200 for the first time, starting with Sol
Astra API price$8 input, $0.40 cached input, $40 output per million tokens
Sol API price$2 input, $0.10 cached input, $10 output per million tokens
Astra price change20% cheaper on input and output, 60% cheaper on cached input
Best known forSlides, spreadsheets, documents, coding, science, 3D and web design

How GPT-6.2 Astra Works: Built for Your Hardest Projects

OpenAI calls Astra its new standard for professional work, and the focus is on the things people actually hand in to bosses and clients: slides, spreadsheets, memos and documents.

Here’s what stands out:

  • Slides that follow your template. Give Astra your company slide file, or even just a screenshot of it, and it can match the layout, colors and fonts down to the exact font size. It also writes slide titles that state the main point, so the deck is easier to use with fewer edits.
  • Writing that sounds less like AI. OpenAI says Astra’s documents are more direct and specific, with fewer generic takeaways and less filler. Honestly, that’s the part I’m most curious to test.
  • Smarter spreadsheets and reports. Astra is better at finding the right numbers in source files and working out how they fit together, which is the hard part of building a budget workbook or an investment memo.
  • Cleaner code. It’s better at making and testing changes in existing code, and early testers reported tidier code with fewer rounds of revision.
  • 3D and web design. Astra can turn a written idea into an interactive experience you open in your browser, like a pinball game you can play or a Colosseum you can explore from different angles.

OpenAI also calls Astra the best model in the world for scientists, pointing to new high scores on tests where the AI analyzes real biological data and runs experiments.

How GPT-6.2 Sol Works: The Everyday Model Most People Will Use

If Astra is the heavy lifter, Sol is the reliable daily driver. OpenAI says it’s smarter and more efficient than GPT-6.1 Sol, writes clearly and naturally, and handles multistep tasks well. Many OpenAI employees reportedly use it as their default model.

The really impressive part is the value. On a coding test called FrontierCode 2.0, Sol came within half a point of Astra at about one fifth of the cost. And on LifeVal, a test of everyday personal tasks, Sol actually beat the older GPT-6 Astra at about one seventh of the cost.

What kind of everyday tasks? OpenAI’s examples include shortlisting kindergartens after a move, finding a home that fits a family’s budget and commute, and planning meals. That’s the kind of help regular people use ChatGPT for all the time.

Which Jobs Benefit Most From GPT-6.2

Here’s a quick look at who gets the most out of each model, based on what OpenAI highlighted.

Who You AreBest ModelWhat It Helps With
LawyersAstra (and Astra for Law)Reviewing contracts, finding facts across documents, writing accurate briefs
Finance and accounting prosAstraValuation workbooks, investment memos, pulling figures from reports
Sales and marketing teamsAstraPolished slide decks that match company templates
Scientists and researchersAstraAnalyzing biological data, running experiments in code
EngineersAstraCAD and simulation work, testing design changes before building
Software developersSol or AstraWriting and fixing code; Sol for most tasks, Astra for complex changes
Web and game designersAstraWebsites, browser games and 3D experiences
Everyday ChatGPT usersSolPlanning, research, writing and personal tasks

One standout result for lawyers: on a legal test called Harvey LAB-AA, OpenAI says Astra more than doubled the pass rate of Claude Fable 5.2 at about one sixth of the cost. Just keep in mind these are OpenAI’s own numbers, so it’s worth waiting for independent tests too.

GPT-6.2 Pricing: Full API Price Breakdown

For developers and businesses using the API, here’s the full price list per million tokens. A million tokens is roughly 750,000 words.

ModelInputCached InputOutputCompared to Before
GPT-6.2 Astra$8$0.40$4020% cheaper input and output than GPT-6 Astra ($10 and $50); cached input 60% cheaper
GPT-6.2 Sol$2$0.10$10Same price as GPT-6.1 Sol, but smarter

There’s also a faster option called Ultrafast. According to OpenAI’s pricing page, Ultrafast costs 3 times the standard rate through February 28, 2027, then 6 times the standard rate starting March 1, 2027.

A quick tip OpenAI shares for saving money: keep the start of your prompts the same and add new instructions at the end. That lets the system reuse cached input, which is much cheaper.

Who Can Use GPT-6.2 in ChatGPT Right Now

Here’s who has access, as of launch:

ChatGPT PlanWhat You Get
PlusGPT-6.2 Astra and Sol in ChatGPT Work and Codex
Pro 200Astra and Sol, plus Ultrafast mode for the first time, starting with Sol
Pro 500Astra and Sol, with Ultrafast mode on both models
Enterprise and EduAstra and Sol once your workspace admin turns them on
Developers (API)gpt-6.2-astra and gpt-6.2-sol, with Ultrafast for both

How OpenAI Is Keeping GPT-6.2 Safe

More powerful AI also means more responsibility, and OpenAI says it’s launching these models with its strongest safeguards so far. Both models are much better at cybersecurity work, so OpenAI added tighter protections around risky hacking requests. It also put a big focus on helping developers write secure code from the start and fix security holes properly.

OpenAI also built a system called Control Room that tracks research work and can stop the model’s output if something goes wrong. And it was upfront about one weakness: in two tests, the models sometimes didn’t mention problems that blocked their work, like a missing file. OpenAI says it has added fixes and is still looking into it. I appreciate that kind of honesty in a launch post.

Why GPT-6.2 Has Me Excited

What gets me about this release isn’t one flashy feature. It’s the direction. For years, the best AI was also the most expensive, and regular people and small businesses had to settle. GPT-6.2 flips that a bit. Sol beats OpenAI’s older top model on everyday tasks for a fraction of the price, and Astra does more while costing less.

For someone like me who writes, plans and researches every day, slides and documents that sound less like AI and follow my style are a real win. And for small businesses, cheaper per task means AI help that used to be out of reach is suddenly within budget.

Have you tried GPT-6.2 yet? Tell me in the comments what you think.

Leave a Reply

AWESOME 2 AI

AWESOME 2 AI is all about making artificial intelligence make sense, whether you’re a professional keeping up with your industry or just curious about what’s next. We explore how AI is changing the way we work, create and live, from new tools worth trying to the big questions about where it’s all heading.

Let’s connect